From cb29d467377c3ad5667f4f161b4a67bee7867e17 Mon Sep 17 00:00:00 2001
From: xiuyechen <xiuyechen@gmail.com>
Date: Wed, 24 Jun 2015 16:19:38 -0400
Subject: [PATCH] v1.1 050415

first version shared with Janelia, with example data provided through
dropbox
---
 BasicPlotMaps.m                               |    9 +
 DrawClusters.m                                |  515 ++-
 DrawClustersOnMap_LSh.m                       |    6 +-
 DrawTiledPics.m                               |   91 +
 GUI_FishExplorer.m                            |  357 +-
 GUIpreload_step1_cleanup.m                    |  268 ++
 GUIpreload_step2_detrend.m                    |   66 +
 GUIpreload_step3.5_PoolClustersFromAllFish.m  |  113 +
 GUIpreload_step3_align.m                      |  298 ++
 GUIpreload_step4_loadGUI.m                    |   30 +
 PartitionForFastSaveLoad.m                    |   85 +
 README.txt                                    |   20 +
 StimulusKeyPreprocessing.m                    |  401 ++
 arrows.jpg                                    |  Bin 0 -> 28083 bytes
 arrows_wide.jpg                               |  Bin 0 -> 25217 bytes
 first version/AccessLocalFunctions.m          |   12 +
 first version/BasicPlotMaps.m                 |  261 ++
 Conv_test.m => first version/Conv_test.m      |    0
 .../DA_2_Solution.m                           |    0
 first version/DrawClusters.m                  |  210 +
 first version/DrawClustersOnMap_LSh.m         |  137 +
 FFT_test.m => first version/FFT_test.m        |    0
 first version/GUI_FishExplorer.m              | 3446 +++++++++++++++++
 first version/GetMotorRegressor.m             |   86 +
 .../GetPhotoTrans.m                           |    0
 first version/GetStimBar.m                    |   25 +
 first version/GetStimRegressor.m              |  186 +
 .../Load_LSh_step1_cleanup.m                  |    0
 .../Load_LSh_step2_detrend.m                  |    0
 .../Load_LSh_step3_align.m                    |    0
 .../Load_LSh_step4_loadGUI.m                  |    0
 Modeling001.m => first version/Modeling001.m  |    0
 .../PlotBy16StimsFromGUI.m                    |    0
 Plotting.m => first version/Plotting.m        |    0
 .../Plotting_withinM.m                        |    0
 .../PoolClustersFromAllFish.m                 |    0
 SortMbystim.m => first version/SortMbystim.m  |    0
 first version/imNormalize99.m                 |   11 +
 first version/push_cIX_gIX.m                  |   10 +
 first version/read_LSstack_fast_float.m       |   15 +
 fishpic.jpg                                   |  Bin 0 -> 25805 bytes
 read_LSstack_fast_float_mex64.mexw64          |  Bin 0 -> 7680 bytes
 42 files changed, 6383 insertions(+), 275 deletions(-)
 create mode 100644 DrawTiledPics.m
 create mode 100644 GUIpreload_step1_cleanup.m
 create mode 100644 GUIpreload_step2_detrend.m
 create mode 100644 GUIpreload_step3.5_PoolClustersFromAllFish.m
 create mode 100644 GUIpreload_step3_align.m
 create mode 100644 GUIpreload_step4_loadGUI.m
 create mode 100644 PartitionForFastSaveLoad.m
 create mode 100644 README.txt
 create mode 100644 StimulusKeyPreprocessing.m
 create mode 100644 arrows.jpg
 create mode 100644 arrows_wide.jpg
 create mode 100644 first version/AccessLocalFunctions.m
 create mode 100644 first version/BasicPlotMaps.m
 rename Conv_test.m => first version/Conv_test.m (100%)
 rename DA_2_Solution.m => first version/DA_2_Solution.m (100%)
 create mode 100644 first version/DrawClusters.m
 create mode 100644 first version/DrawClustersOnMap_LSh.m
 rename FFT_test.m => first version/FFT_test.m (100%)
 create mode 100644 first version/GUI_FishExplorer.m
 create mode 100644 first version/GetMotorRegressor.m
 rename GetPhotoTrans.m => first version/GetPhotoTrans.m (100%)
 create mode 100644 first version/GetStimBar.m
 create mode 100644 first version/GetStimRegressor.m
 rename Load_LSh_step1_cleanup.m => first version/Load_LSh_step1_cleanup.m (100%)
 rename Load_LSh_step2_detrend.m => first version/Load_LSh_step2_detrend.m (100%)
 rename Load_LSh_step3_align.m => first version/Load_LSh_step3_align.m (100%)
 rename Load_LSh_step4_loadGUI.m => first version/Load_LSh_step4_loadGUI.m (100%)
 rename Modeling001.m => first version/Modeling001.m (100%)
 rename PlotBy16StimsFromGUI.m => first version/PlotBy16StimsFromGUI.m (100%)
 rename Plotting.m => first version/Plotting.m (100%)
 rename Plotting_withinM.m => first version/Plotting_withinM.m (100%)
 rename PoolClustersFromAllFish.m => first version/PoolClustersFromAllFish.m (100%)
 rename SortMbystim.m => first version/SortMbystim.m (100%)
 create mode 100644 first version/imNormalize99.m
 create mode 100644 first version/push_cIX_gIX.m
 create mode 100644 first version/read_LSstack_fast_float.m
 create mode 100644 fishpic.jpg
 create mode 100644 read_LSstack_fast_float_mex64.mexw64

diff --git a/BasicPlotMaps.m b/BasicPlotMaps.m
index 9196801..f8e5c37 100644
--- a/BasicPlotMaps.m
+++ b/BasicPlotMaps.m
@@ -18,6 +18,15 @@ function BasicDrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,is
 barratio = 0.02;
 numK = double(max(double(numK),double(max(gIX))));
 
+% down-sample
+displaymax = 1000;
+numcell = size(M,1);
+if numcell > displaymax,
+    skip = round(numcell/displaymax);
+    M = M(1:skip:end,:);
+    gIX = gIX(1:skip:end,:);
+end
+
 % sort traces by index
 im = M;
 [nLines,nFrames] = size(im);
diff --git a/DrawClusters.m b/DrawClusters.m
index 964da88..524ce77 100644
--- a/DrawClusters.m
+++ b/DrawClusters.m
@@ -1,201 +1,293 @@
-function DrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,iswrite)
-pos = get(gca,'Position');
-barratio = 0.03;
+function DrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,iswrite,isPopout,isPlotLines,isPlotFictive)
+axis off;
+pos = get(h1,'Position'); % [left, bottom, width, height]    
 
-%% Prepare cluster data
-% down-sample
-displaymax = 1000;
-numcell = size(M,1);
-if numcell > displaymax,
-    skip = round(numcell/displaymax);
-    M = M(1:skip:end,:);
-    gIX = gIX(1:skip:end,:);
-end
-
-numK = double(max(double(numK),double(max(gIX))));
+if isPlotLines,
+    if 1, % (just to collapse isPlotLines = true)
+        %% settings
+        len = pos(3);
+        fps = 1.97;
+        
+        % set position grid
+        nLines = length(unique(gIX));
+        if isPlotFictive,
+            nExtraRows = 3; % number of extra rows
+        else
+            nExtraRows = 2;
+        end        
+        nRows = max(8,nLines)+nExtraRows;
+        lineH = pos(4)/nRows;
+%         Lpos = [pos(1),pos(2)+pos(4)-lineH*nRows,pos(3),lineH*nRows];
+% %         Lpos = [pos(1),pos(2)+pos(4)-lineH*(nLines+nExtraRows),pos(3),lineH*(nLines+nExtraRows)];
+%         set(h1,'Position',Lpos);
 
-% sort traces by index
-im = M;
-[nLines,nFrames] = size(im);
-[~,I] = sort(gIX);
-im = im(I,:);
+        %     figure('Position',[500,900-(min(nLines,15)+nDiv)*50,600,(min(nLines,15)+nDiv)*50],'color',[1 1 1]);
+        xv = 1:size(M,2);        
+                
+        %% Set colormap        
+        if strcmp(clrmap,'jet'),
+            clr = flipud(jet(double(numK)))*0.8; % make darker by *0.8
+        else % if strcmp(clrmap,'hsv'), 
+            clr = hsv(round(double(numK)*1.1))*0.7; % make darker by *0.7
+        end
+        % clr = lines(nLines);
+        % clr = ones(nLines,3)*0.3; % uniform charcoal
+        
+        %% draw stimulus bar
+        stimbar = GetStimBar(1,stim);
+        rpos = [pos(1),pos(2)+pos(4)-3/4*lineH,len,lineH/2];
+        h = axes('Position',rpos);
+        
+        image(stimbar);
+        set(h,'Xtick',[],'Ytick',[])
+        
+        % draw fish icon
+        rpos = [pos(1)*0.2,pos(2)+pos(4)-3/4*lineH,pos(1)*0.7,lineH/2];
+        axes('Position',rpos);
+        DrawFishIcon;    
+                
+        %% draw cluster means
+        Ymeans = zeros(nLines,size(M,2));
+        U = unique(gIX);
+        for j = 1:nLines,
+            k = U(j);
+            %     ymean = Cz(k,:);
+            Ys = M(gIX==k,:);
+            ymean = mean(Ys,1);
+            Ymeans(k,:) = ymean;
+            rpos = [pos(1),pos(2)+pos(4)-lineH*(j+1),len,lineH];
+            axes('Position',rpos); hold on;
+            
+            ySTD=std(Ys); % /sqrt(size(Ys,1))
+            ySTD_upper=ymean+ySTD;
+            ySTD_lower=ymean-ySTD;
+            h = fill([xv fliplr(xv)], [ySTD_upper fliplr(ySTD_lower)],0.5+0.3*clr(k,:));
+            set(h, 'EdgeColor','none')
+            
+            plot(xv,ymean,'-','Linewidth',1,'color',clr(k,:))
+            axis tight;axis off
+        end
+        
+        %% plot scale bar
+        axes('Position',[pos(1),pos(2)+pos(4)-lineH*(nLines+2),len,lineH]); hold on;
+        if size(M,2)<1200, % <~10 min, plot 1min scale bar
+            plot([1,60*fps],[0.75,0.75],'k-');
+            text(pos(1)+60*fps/2,0.5,'1 min','HorizontalAlignment','center')
+        else % >~10 min, plot 10min scale bar
+            plot([1,600*fps],[0.75,0.75],'k-');
+            text(pos(1)+600*fps/2,0.5,'10 min','HorizontalAlignment','center')
+        end
+        xlim([xv(1),xv(end)]);ylim([0,1]);
+        axis off
+                
+        %% draw fictive behavior bar
+        if isPlotFictive,
+            h = axes('Position',[pos(1),pos(2)+pos(4)-lineH*(nLines+3),len,0.7/(nLines+nExtraRows)]);
+            DrawFictiveBar(h,fictive);
+            
+            axes('Position',[0.02,pos(2)+pos(4)-lineH*(nLines+3),pos(1)-0.02,0.7/(nLines+nExtraRows)]);
+            DrawArrowsIcon(isPopout);
+        end
 
-%% convert imagesc effect into rgb matrix 'RGB'
-if dataFR, % is drawing cell-response fluorescence
-    cmap = gray(64);
-    im = AutoScaleImage0to1(im); % scaling to min/max 0/1
-    minlim = 0;
-    maxlim = 1;
-else % create blue-white-red map, for regression results
-    cmap = zeros(64,3);
-    cmap(:,1) = [linspace(0,1,32), linspace(1,1,32)];
-    cmap(:,2) = [linspace(0,1,32), linspace(1,0,32)];
-    cmap(:,3) = [linspace(1,1,32), linspace(1,0,32)];
-    minlim = -1; %min(min(im));
-    maxlim = 1; %max(max(im));
-end
-RGB = ImageToRGB(im,cmap,minlim,maxlim); % map image matrix to range of colormap
+    end
+    
+else
 
-%% add vertical color code
-if exist('clrmap','var'),
-    if strcmp(clrmap,'jet'),
-        temp = flipud(jet(numK));
+    if ~isPopout,
+        barratio = 0.025;
+    else
+        barratio = 0.05;
+    end
+    %% Prepare cluster data
+    % down-sample
+    displaymax = 1000;
+    numcell = size(M,1);
+    if numcell > displaymax,
+        skip = round(numcell/displaymax);
+        M = M(1:skip:end,:);
+        gIX = gIX(1:skip:end,:);
+    end
+    
+%     numK = double(max(double(numK),double(max(gIX))));
+    
+    % sort traces by index
+    im = M;
+    [nLines,nFrames] = size(im);
+    [~,I] = sort(gIX);
+    im = im(I,:);
+    
+    %% convert imagesc effect into rgb matrix 'RGB'
+    if dataFR, % is drawing cell-response fluorescence
+        cmap = gray(64);
+        im = AutoScaleImage0to1(im); % scaling to min/max 0/1
+        minlim = 0;
+        maxlim = 1;
+    else % create blue-white-red map, for regression results
+        cmap = zeros(64,3);
+        cmap(:,1) = [linspace(0,1,32), linspace(1,1,32)];
+        cmap(:,2) = [linspace(0,1,32), linspace(1,0,32)];
+        cmap(:,3) = [linspace(1,1,32), linspace(1,0,32)];
+        minlim = -1; %min(min(im));
+        maxlim = 1; %max(max(im));
+    end
+    RGB = ImageToRGB(im,cmap,minlim,maxlim); % map image matrix to range of colormap
+    
+    %% add vertical color code
+    if exist('clrmap','var'),
+        if strcmp(clrmap,'jet'),
+            temp = flipud(jet(numK));
+        else % 'hsv'
+            temp = hsv(round(numK*1.1));
+        end
     else % 'hsv'
         temp = hsv(round(numK*1.1));
     end
-else % 'hsv'
-    temp = hsv(round(numK*1.1));
-end
-cmap2 = vertcat(temp(1:numK,:),[0,0,0]); % extend colormap to include black
-bwidth = max(round(nFrames/30),1);
-
-idx = gIX(I);
-ix_div = [find(diff(idx));length(idx)];
-
-bars = ones(nLines,bwidth);
-if numK>1,
-    bars(1:ix_div(1),:) = idx(ix_div(1));
-    for i = 2:length(ix_div),
-        % paint color
-        bars(ix_div(i-1)+1:ix_div(i),:) = idx(ix_div(i));
-    end
-end
-im_bars = reshape(cmap2(bars,:),[size(bars),3]);
-% add a white division margin
-div = ones(nLines,round(bwidth/2),3);
-% put 3 parts together
-im = horzcat(RGB,div,im_bars);
-
-%% plot figure
-
-[s1,s2,s3] = size(im);
-
-hold off;
-set(h1,'Position',[pos(1),pos(2)+pos(4)*barratio*3,pos(3),pos(4)*(1-4*barratio)]); 
-if s1<30,
-    temp = im;
-    im = ones(30,s2,s3);
-    im(1:s1,:,:) = temp;
-end
-image(im);
-set(gca, 'box', 'off')
-
-hold on;
-
-% plot cluster division lines
-plot([0.5,s2+0.5],[0.5,0.5],'k','Linewidth',0.5);
-if numK>1,
-    for i = 1:length(ix_div),% = numK-1,
-        y = ix_div(i)+0.5;
-        plot([0.5,s2+0.5],[y,y],'k','Linewidth',0.5);
+    cmap2 = vertcat(temp(1:numK,:),[0,0,0]); % extend colormap to include black
+    bwidth = max(round(nFrames/30),1);
+    
+    idx = gIX(I);
+    ix_div = [find(diff(idx));length(idx)];
+    
+    bars = ones(nLines,bwidth);
+    if numK>1,
+        bars(1:ix_div(1),:) = idx(ix_div(1));
+        for i = 2:length(ix_div),
+            % paint color
+            bars(ix_div(i-1)+1:ix_div(i),:) = idx(ix_div(i));
+        end
     end
-end
-
-ylabel(['Number of cells: ' num2str(numcell)]);
-set(gca,'YTick',[],'XTick',[]);
-set(gcf,'color',[1 1 1]);
-
-colormap gray;
-
-% write in number label
-x = s2*1.003;
-y_last = 0;
-for i = 1:length(ix_div),
-    % avoid label crowding
-    margin = 0.015*s1;
-    y0 = ix_div(i)+0.5;
-    y = max(y_last+margin,y0);
-    if i<length(ix_div),
-        ynext = ix_div(i+1);
+    im_bars = reshape(cmap2(bars,:),[size(bars),3]);
+    % add a white division margin
+    div = ones(nLines,round(bwidth/2),3);
+    % put 3 parts together
+    im = horzcat(RGB,div,im_bars);
+    
+    %% plot figure
+    
+    [s1,s2,s3] = size(im);
+    
+    hold off;
+    if ~isPlotFictive,
+        set(h1,'Position',[pos(1),pos(2),pos(3),pos(4)*(1-barratio)]);
     else
-        ynext = y0+margin*2;
+        set(h1,'Position',[pos(1),pos(2)+pos(4)*barratio*2,pos(3),pos(4)*(1-barratio*3)]);
     end
-    % draw if not squeezed too much away
-    if y < y0+margin*2 && y < 0.5*(y0+ynext), %nLines+stimheight-margin,
-        if exist('iswrite','var') && iswrite,
-            try
-                text(x,y,[num2str(idx(ix_div(i))) ': ' num2str(rankscore(i))],'HorizontalAlignment','Left','VerticalAlignment','Bottom',...
-                    'FontUnits','normalized','FontSize',0.015,'Tag','label');
-            catch
-                disp('rankscore invalid');
+    % axis off
+    if s1<30,
+        temp = im;
+        im = ones(30,s2,s3);
+        im(1:s1,:,:) = temp;
+    end
+    image(im);
+    set(gca, 'box', 'off')
+    
+    hold on;
+    
+    % plot cluster division lines
+    plot([0.5,s2+0.5],[0.5,0.5],'k','Linewidth',0.5);
+    if numK>1,
+        for i = 1:length(ix_div),% = numK-1,
+            y = ix_div(i)+0.5;
+            plot([0.5,s2+0.5],[y,y],'k','Linewidth',0.5);
+        end
+    end
+    
+    ylabel(['Number of cells: ' num2str(numcell)]);
+    set(gca,'YTick',[],'XTick',[]);
+    set(gcf,'color',[1 1 1]);
+    
+    colormap gray;
+    
+    % write in number label
+    x = s2*1.003;
+    y_last = 0;
+    for i = 1:length(ix_div),
+        % avoid label crowding
+        margin = 0.015*s1;
+        y0 = ix_div(i)+0.5;
+        y = max(y_last+margin,y0);
+        if i<length(ix_div),
+            ynext = ix_div(i+1);
+        else
+            ynext = y0+margin*2;
+        end
+        % draw if not squeezed too much away
+        if y < y0+margin*2 && y < 0.5*(y0+ynext), %nLines+stimheight-margin,
+            if exist('iswrite','var') && iswrite,
+                try
+                    text(x,y,[num2str(idx(ix_div(i))) ': ' num2str(rankscore(i))],'HorizontalAlignment','Left','VerticalAlignment','Bottom',...
+                        'FontUnits','normalized','FontSize',0.015,'Tag','label');
+                catch
+                    disp('rankscore invalid');
+                    text(x,y,num2str(idx(ix_div(i))),'HorizontalAlignment','Left','VerticalAlignment','Bottom',...
+                        'FontUnits','normalized','FontSize',0.015,'Tag','label');
+                end
+                
+            else
                 text(x,y,num2str(idx(ix_div(i))),'HorizontalAlignment','Left','VerticalAlignment','Bottom',...
                     'FontUnits','normalized','FontSize',0.015,'Tag','label');
             end
-            
-        else
-            text(x,y,num2str(idx(ix_div(i))),'HorizontalAlignment','Left','VerticalAlignment','Bottom',...
-                'FontUnits','normalized','FontSize',0.015,'Tag','label');
+            y_last = y;
         end
-        y_last = y;
     end
-end
-
-%% Draw stim bars
-halfbarheight = 100;
-stimbar_ = GetStimBar(halfbarheight,stim);
-
-% pad with margin on right to fit combo-image size
-barlength = size(im,2);
-stimbar = ones(halfbarheight*2,barlength,3);
-stimbar(:,1:nFrames,:) = stimbar_(:,1:nFrames,:); % crop to size
-
-% top axis
-axes('Position',[pos(1),pos(2)+pos(4)*(1-barratio),pos(3),pos(4)*barratio]);
-image(stimbar);axis off;hold on;
-plot([1,length(stim)],[1,1],'k','Linewidth',0.5); % plot top border
-plot([0.5,0.5],[0,size(stimbar,1)],'k','Linewidth',0.5); % plot left/right border
-plot([length(stim),length(stim)],[0,size(stimbar,1)],'k','Linewidth',0.5);
-
-% bottom axis
-axes('Position',[pos(1),pos(2)+pos(4)*barratio*2,pos(3),pos(4)*barratio]);
-image(stimbar);axis off;hold on;
-plot([1,length(stim)],[size(stimbar,1),size(stimbar,1)],'k','Linewidth',0.5); % plot bottom border
-plot([0.5,0.5],[0,size(stimbar,1)],'k','Linewidth',0.5); % plot left/right border
-plot([length(stim),length(stim)],[0,size(stimbar,1)],'k','Linewidth',0.5);
-
-%% Draw fictive
-axes('Position',[pos(1),pos(2),pos(3),pos(4)*barratio*2]);
-
-temp = ones(size(fictive,1),barlength);
-temp(:,1:length(fictive)) = fictive;
-imagesc(temp);colormap hot
-set(gca,'YTick',[],'XTick',[]);
-set(gcf,'color',[1 1 1]);
-set(gca, 'box', 'off')
-hold on;axis ij;
-
-% plot division lines
-for i = 0:3,
-    y = i+0.5;
-    plot([0.5,length(fictive)+0.5],[y,y],'w','Linewidth',0.5);
-end
-% labels
-names = {'Left','Right','Forward','Raw L','Raw R'};
-x = -s2*0.05;
-for i = 1:5,
-    y = i;
-    text(x,y,names{i},'Fontsize',7);
-end
-
-%% horizontal axis: approximate time length, assuming 2 fps (real is 1.97 for now)
-set(gca,'XColor',[1 1 1]);
-if s2/2/60<5, % <5 min
-    l = xlabel(['Length of display: ~ ' num2str(s2/2) 'sec'],'Color',[0 0 0]);
-else
-    l = xlabel(['Length of display: ~ ' num2str(s2/2/60) 'min'],'Color',[0 0 0]);
-end
+    
+    %% Draw stim bars
+    halfbarheight = 100;
+    stimbar_ = GetStimBar(halfbarheight,stim);
+    
+    % pad with margin on right to fit combo-image size
+    barlength = size(im,2);
+    stimbar = ones(halfbarheight*2,barlength,3);
+    stimbar(:,1:nFrames,:) = stimbar_(:,1:nFrames,:); % crop to size
+    
+    % top axis
+    
+    axes('Position',[pos(1),pos(2)+pos(4)*(1-barratio),pos(3),pos(4)*barratio]);
+    image(stimbar);axis off;hold on;
+    plot([1,length(stim)],[1,1],'k','Linewidth',0.5); % plot top border
+    plot([0.5,0.5],[0,size(stimbar,1)],'k','Linewidth',0.5); % plot left/right border
+    plot([length(stim),length(stim)],[0,size(stimbar,1)],'k','Linewidth',0.5);
+    
+    % bottom axis
+    if ~isPopout,
+        axes('Position',[pos(1),pos(2)+pos(4)*barratio*2,pos(3),pos(4)*barratio]);
+        image(stimbar);axis off;hold on;
+        plot([1,length(stim)],[size(stimbar,1),size(stimbar,1)],'k','Linewidth',0.5); % plot bottom border
+        plot([0.5,0.5],[0,size(stimbar,1)],'k','Linewidth',0.5); % plot left/right border
+        plot([length(stim),length(stim)],[0,size(stimbar,1)],'k','Linewidth',0.5);
+    end
+    %% Draw fictive
+    if isPlotFictive,
+        h = axes('Position',[pos(1),pos(2),pos(3),pos(4)*barratio*2]);        
+        DrawFictiveBar(h,fictive,barlength);        
+        axes('Position',[0.015,pos(2),0.03,pos(4)*barratio*2]);
+        DrawArrowsIcon(isPopout);
+    else
+        h = h1;
+    end
 
-%% Draw fish icon:
-if exist('fishpic.jpg','file')==2,
-    fishpic = imread('fishpic.jpg');
-    axes('Position',[0.015, 0.095, 0.03, 0.02]);
-    image(fishpic); axis off; axis equal
+    % label horizontal axis: approximate time length, assuming 1.97 fps
+    set(h,'XColor',[1 1 1]);
+    if s2/2/60<5, % <5 min
+        xlabel(['Time ~ ' num2str(round(s2/1.97)) 'sec'],'Color',[0 0 0]);
+    else
+        xlabel(['Time ~ ' num2str(round(s2/1.97/60)) 'min'],'Color',[0 0 0]);
+    end
+    
+    %% Draw fish icon:
+    if ~isPopout,
+        axes('Position',[0.015, pos(2)+pos(4)*barratio*2, 0.03, pos(4)*barratio]);
+    else
+        axes('Position',[0.015,pos(2)+pos(4)*(1-barratio),0.03,pos(4)*barratio]);
+    end
+    DrawFishIcon;
 end
 end
 
 %% Internal Functions
 
 function im = AutoScaleImage0to1(im)
+im(isnan(im)) = 0;
 temp = mat2gray(im);
 low_high = stretchlim(temp,[0.005,0.995]);
 temp(temp<low_high(1)) = low_high(1);
@@ -208,3 +300,74 @@ function DrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,iswrite
 ix = round(interp1(linspace(minlim,maxlim,L),1:L,im,'linear','extrap'));
 RGB = reshape(cmap(ix,:),[size(ix) 3]); % Make RGB image from scaled.
 end
+
+function DrawFishIcon
+if exist('fishpic.jpg','file')==2,
+    pic = imread('fishpic.jpg');
+    fishpic = imresize(pic,0.5);
+    
+    image(fishpic); axis off; axis equal; axis tight
+end
+end
+
+function DrawArrowsIcon(isPopout)
+if ~isPopout,
+    if exist('arrows.jpg','file')==2,
+        pic = imread('arrows.jpg');
+        arrowpic = imresize(pic,0.2);
+        image(arrowpic); axis off; axis image; axis tight
+    end
+else
+    if exist('arrows_wide.jpg','file')==2,
+        pic = imread('arrows_wide.jpg');
+        arrowpic = imresize(pic,0.2);
+        image(arrowpic); axis off; axis image; axis tight
+    end
+end
+end
+
+function DrawFictiveBar(h,fictive,barlength)
+regressors = GetMotorRegressor(fictive);
+fc = reshape([regressors(1:5).im],length(fictive),[])';
+fc = AutoScaleImage0to1(fc);
+if exist('barlength','var'),
+    temp = ones(size(fc,1),barlength);
+    temp(:,1:length(fc)) = fc;
+else
+    temp = fc;
+end
+if 0, % plot all 5 lines
+%     imagesc(temp);colormap gray
+%     set(h,'YTick',[],'XTick',[]);
+%     %     set(gcf,'color',[1 1 1]);
+%     set(h, 'box', 'off')
+%     hold on;axis ij;
+%     
+%     % plot division lines
+%     for i = 0:3,
+%         y = i+0.5;
+%         plot([0.5,length(fictive)+0.5],[y,y],'w','Linewidth',0.5);
+%     end
+%     % labels
+%     names = {'Left','Right','Forward','Raw L','Raw R'};
+%     x = -s2*0.05;
+%     for i = 1:5,
+%         y = i;
+%         text(x,y,names{i},'Fontsize',7);
+%     end
+    
+else % only plot top 3 lines
+    fc = vertcat(temp(1,:),temp(3,:),temp(2,:));
+    imagesc(fc);colormap gray
+    set(h,'YTick',[],'XTick',[]);
+    %     set(gcf,'color',[1 1 1]);
+    set(h, 'box', 'off')
+    hold on;axis ij;
+    
+    % plot division lines
+    for i = 0:2,
+        y = i+0.5;
+        plot([0.5,length(fictive)+0.5],[y,y],'w','Linewidth',0.5);
+    end    
+end
+end
diff --git a/DrawClustersOnMap_LSh.m b/DrawClustersOnMap_LSh.m
index 5bbeb48..6ca10bf 100644
--- a/DrawClustersOnMap_LSh.m
+++ b/DrawClustersOnMap_LSh.m
@@ -27,11 +27,11 @@
 end
 
 if strcmp(clrmap,'jet'),
-    temp = flipud(jet(numK));
+    cmap = flipud(jet(numK));
 else % 'hsv'
-    temp = hsv(round(numK*1.1));
+    n = round(numK*1.1);
+    cmap = hsv(max(1,n));
 end
-cmap = temp(1:numK,:); % extend colormap to include black
 
 anat_YX = anat_yx/4;
 anat_YZ = anat_yz/4;
diff --git a/DrawTiledPics.m b/DrawTiledPics.m
new file mode 100644
index 0000000..41a22ab
--- /dev/null
+++ b/DrawTiledPics.m
@@ -0,0 +1,91 @@
+function DrawTiledPics(hfig)
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+CInfo = getappdata(hfig,'CInfo');
+ave_stack = getappdata(hfig,'ave_stack');
+
+timestamp = datestr(now,'mmddyy_HHMMSS');
+tiffName = ['stack_' timestamp '.tif'];
+
+U = unique(gIX);
+numK = length(U);
+
+% left half: stack with cells marked;
+% right half: original anatomy, or mark all cells
+
+ave_stack2=zeros(size(ave_stack,1), size(ave_stack,2), size(ave_stack,3) ,3);
+nPlanes=size(ave_stack,3);
+dimv_yxz=size(ave_stack);
+stacklen=numel(ave_stack);
+
+% circle=makeDisk2(10,21);
+radius = 10; dim = 21;
+center=floor(dim/2)+1;
+circle=zeros(dim);
+for x=1:dim
+    for y=1:dim
+        if norm([x,y]-[center,center])<=radius
+            circle(x,y)=1;
+        end
+    end
+end
+
+
+[r, v]=find(circle);
+r=r-11;v=v-11;
+circle_inds  = r*dimv_yxz(1)+v;
+cmap = hsv(round(numK*1.1));
+% cmap = [0.3 1 0];
+weight = 0.3;
+
+for i=1:nPlanes,
+    ave_stack2(:,:,i,:)=repmat(imNormalize99(ave_stack(:,:,i))/4,[1 1 1 3]);
+end
+
+for j=1:length(cIX)
+    cinds=(CInfo(cIX(j)).center(2)-1)*dimv_yxz(1)+CInfo(cIX(j)).center(1);
+    labelinds=find((cinds+circle_inds)>0 & (cinds+circle_inds)<=dimv_yxz(1)*dimv_yxz(2));
+    zinds=dimv_yxz(1)*dimv_yxz(2)*(CInfo(cIX(j)).slice-1);
+    ix = find(U==gIX(j));
+    ixs = cinds+circle_inds(labelinds)+zinds;
+    ave_stack2(ixs)=cmap(ix,1)*weight + ave_stack2(ixs)*(1-weight);
+    ixs = cinds+circle_inds(labelinds)+zinds+stacklen;
+    ave_stack2(ixs)=cmap(ix,2)*weight + ave_stack2(ixs)*(1-weight);
+    ixs = cinds+circle_inds(labelinds)+zinds+stacklen*2;
+    ave_stack2(ixs)=cmap(ix,3)*weight + ave_stack2(ixs)*(1-weight);
+end
+
+%% view stack sequentially
+% figure;
+% for i_plane = 1:nPlanes,
+%     im = squeeze(ave_stack2(:,:,i_plane,:));
+%     image(im);
+%     axis image; axis off
+%     pause(0.1)
+% end
+
+% lay out
+
+h = figure('Position',[10,20,round(1136*1.4),round(683*1.4)],'color','k');
+hold on;
+
+axes('Position',[0,0,1,1]); % BS fix to get the figure background to stay when saving pics
+image(zeros(3,3,3));axis off
+
+numRow = 3;
+numP = nPlanes - mod(nPlanes,numRow);
+numCol = numP/numRow;
+
+for i_plane = 1:numP,
+    im = squeeze(ave_stack2(:,:,i_plane,:));
+    im = imresize(im,0.25);
+    
+    [col, row] = ind2sub([numCol,numRow],i_plane); % this is intensionally inverted...
+    posVec = [(col-1)*1/numCol,1-row*1/numRow,1/numCol,1/numRow];
+    axes('Position',posVec);
+    image(im);
+    axis image; axis off
+    %     pause(0.1)
+end
+
+end
\ No newline at end of file
diff --git a/GUI_FishExplorer.m b/GUI_FishExplorer.m
index a417875..04405e0 100644
--- a/GUI_FishExplorer.m
+++ b/GUI_FishExplorer.m
@@ -21,9 +21,10 @@
 % global VAR;
 % load('VAR_current.mat','VAR');
 % load('CONSTs_current.mat','CONSTs');
-% hfig = GUI_FishExplorer(CONSTs,VAR);
+% data_dir = [pwd '\example data'];
+% [hfig,fcns] = GUI_FishExplorer(data_dir,CONSTs,VAR);
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-% Full datasets are stored as 'CONST_F?.mat' (? = fish number), to load from the GUI
+% Full datasets are stored as 'CONST_F?_fast.mat' (? = fish number), to load from the GUI
 
 %% User Interface:
 % function hfig = GUI_FishExplorer(CONSTs,VAR,CONST,i_fish)
@@ -57,7 +58,11 @@
 %% Pass external variables into appdata (stored with main figure handle)
 setappdata(hfig,'CONSTs',CONSTs);
 setappdata(hfig,'VAR',VAR);
-nFish = length(CONSTs);
+nFish = length(CONSTs) + 1; %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+
+% fish protocol sets (different sets have different parameters)
+M_fish_set = [1, 1, 1, 1, 1, 1, 1, 2, 3, 3]; % M = Matrix
+setappdata(hfig,'M_fish_set',M_fish_set);
 
 % parameters / constants
 setappdata(hfig,'z_res',19.7); % resoltion in z, um per slice
@@ -81,9 +86,6 @@
 QuickUpdateFish(hfig,i_fish,'init');
 
 %% Initialize internal params into appdata
-% fish protocol sets (different sets have different parameters)
-M_fish_set = [1, 1, 1, 1, 1, 1, 1, 2, 1]; % M = Matrix
-setappdata(hfig,'M_fish_set',M_fish_set);
 
 % thresholds
 thres_merge = 0.9;
@@ -100,6 +102,8 @@
 setappdata(hfig,'clrmap','hsv');
 setappdata(hfig,'opID',0);
 setappdata(hfig,'rankID',0); 
+setappdata(hfig,'isPlotLines',0); 
+setappdata(hfig,'isPlotFictive',1); 
 setappdata(hfig,'rankscore',[]);
 setappdata(hfig,'isCentroid',0);
 setappdata(hfig,'isWkmeans',1); % in autoclustering, with/without kmeans
@@ -148,7 +152,7 @@
 
 % various UI element handles ('global' easier than passing around..)
 global hback hfwd hclassname hclassmenu hclusgroupmenu hclusmenu hclusname...
-    hdatamenu hopID hcentroid hdataFR hloadfish hfishnum hstimreg hmotorreg...
+    hdatamenu hopID hdataFR hloadfish hfishnum hstimreg hmotorreg...
     hcentroidreg;
 
 %% UI ----- tab one ----- (General)
@@ -176,12 +180,30 @@
     'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
     'Callback',@pushbutton_writeZstack_Callback);
 
+i=i+n;
+n=2; % plots selected cells on anatomy z-stack, tiled display
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Tile Zstack',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_tileZstack_Callback);
+
 i=i+n;
 n=2; % make new figure without the GUI components, can save manually from default menu
 uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Popup plot',...
     'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
     'Callback',@pushbutton_popupplot_Callback);
 
+i=i+n;
+n=2; % popupplot option: whether to plot cluster mean lines instead of all raw traces
+uicontrol('Parent',tab{i_tab},'Style','checkbox','String','Plot lines',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@checkbox_isPlotLines_Callback);
+
+i=i+n;
+n=2; % popupplot option: whether to plot fictive behavior bar
+uicontrol('Parent',tab{i_tab},'Style','checkbox','String','Plot fictive','Value',1,...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@checkbox_isPlotFictive_Callback);
+
 i=i+n;
 n=3; % export main working variables to workspace, can customize!
 uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Export to workspace',...
@@ -233,7 +255,7 @@
 
 i=i+n;
 n=4; % only centroids (~mean) of clusters shown on left-side plot, the rest is unchanged
-hcentroid = uicontrol('Parent',tab{i_tab},'Style','checkbox','String','Display centroids of clusters',...
+uicontrol('Parent',tab{i_tab},'Style','checkbox','String','Display centroids of clusters',...
     'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
     'Callback',@checkbox_showcentroids_Callback);
 
@@ -413,7 +435,7 @@
 
 i=i+n;
 n=2; % (unlike stim regressors, names hardcoded, not importet from regressor...)
-menu = {'(choose)','right swims','left swims','forward swims','raw right','raw left','raw average'};
+menu = {'(choose)','left swims','right swims','forward swims','raw left','raw right','raw average'};
 hmotorreg = uicontrol('Parent',tab{i_tab},'Style','popupmenu','String',menu,'Value',1,...
     'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
     'Callback',@popup_getmotorreg_Callback);
@@ -493,19 +515,19 @@
     'Callback',{@pushbutton_IterCentroidRegression_Callback});
 
 %% UI row 3: (TBD)
-i_row = 3; % old idea: display correlation values of multiple regressors (instead of fluo. trace)
-i = 1;n = 0; % and then can cluster that and further manipulate... 
-
-i=i+n;
-n=4; % did not implement combination of regressors in this version... but display option is coded (dataFR==0)
-uicontrol('Parent',tab{i_tab},'Style','text','String','regressor combos??(TBD)',... 
-    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
-
-i=i+n;
-n=4; % also data storage is coded, 'M' could be loaded from M_0_fluo or M_0_reg (storing reg corr values)
-hdataFR = uicontrol('Parent',tab{i_tab},'Style','checkbox','String','Display fluo./regression',...
-    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
-    'Callback',@checkbox_dataFR_Callback);
+% i_row = 3; % old idea: display correlation values of multiple regressors (instead of fluo. trace)
+% i = 1;n = 0; % and then can cluster that and further manipulate... 
+% 
+% i=i+n;
+% n=4; % did not implement combination of regressors in this version... but display option is coded (dataFR==0)
+% uicontrol('Parent',tab{i_tab},'Style','text','String','regressor combos??(TBD)',... 
+%     'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+% 
+% i=i+n;
+% n=4; % also data storage is coded, 'M' could be loaded from M_0_fluo or M_0_reg (storing reg corr values)
+% hdataFR = uicontrol('Parent',tab{i_tab},'Style','checkbox','String','Display fluo./regression',...
+%     'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+%     'Callback',@checkbox_dataFR_Callback);
 
 %% UI ----- tab four ----- (Clustering etc.)
 i_tab = 4;
@@ -977,6 +999,17 @@ function pushbutton_writeZstack_Callback(hObject,~)
 close(h)
 end
 
+function pushbutton_tileZstack_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+isfullfish = getappdata(hfig,'isfullfish');
+if isfullfish,
+    disp('Rendering...')
+    DrawTiledPics(hfig);
+else
+    errordlg('Load full fish first!');
+end
+end
+
 function out = makeDisk2(radius, dim)
 center=floor(dim/2)+1;
 out=zeros(dim);
@@ -997,7 +1030,7 @@ function pushbutton_popupplot_Callback(hObject,~)
 % figure('Position',[50,100,853,512]); % right plot size: 2048x1706
 % h1 = axes('Position',[0, 0, 0.5, 1]); % left ~subplot
 % h2 = axes('Position',[0.50, 0, 0.5, 1]); % right ~subplot
-figure('Position',[50,100,1600,900]);
+figure('Position',[50,100,1700,900],'color',[1 1 1]);
 h1 = axes('Position',[0.05, 0.03, 0.53, 0.94]); % left ~subplot
 h2 = axes('Position',[0.61, 0.03, 0.35, 0.94]); % right ~subplot
 
@@ -1017,22 +1050,39 @@ function pushbutton_popupplot_Callback(hObject,~)
 clrmap = getappdata(hfig,'clrmap');
 rankscore = getappdata(hfig,'rankscore');
 rankID = getappdata(hfig,'rankID');
-
 iswrite = (rankID>=2);
+isPlotLines = getappdata(hfig,'isPlotLines');
+isPlotFictive = getappdata(hfig,'isPlotFictive');
+
+isPopout = 1;
 
 % left subplot
 axes(h1);
 if isCentroid,
     [C,~] = FindCentroid(gIX,M);
-    DrawClusters(h1,C,unique(gIX),dataFR,numK,stim,fictive,clrmap,rankscore,iswrite);
+    DrawClusters(h1,C,unique(gIX),dataFR,numK,stim,fictive,clrmap,rankscore,...
+        iswrite,isPopout,isPlotLines,isPlotFictive);
+%     DrawClusters(h1,C,unique(gIX),dataFR,numK,stim,fictive,clrmap,rankscore,iswrite,ispopout);
 else
-    DrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,iswrite);
+    DrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,...
+        iswrite,isPopout,isPlotLines,isPlotFictive);
+%     DrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,iswrite,ispopout);
 end
 
 % right subplot
 axes(h2);
-DrawClustersOnMap_LSh(CInfo,cIX,gIX,numK,anat_yx,anat_yz,anat_zx,clrmap);
+DrawClustersOnMap_LSh(CInfo,cIX,gIX,numK,anat_yx,anat_yz,anat_zx,clrmap,'full');
+
+end
+
+function checkbox_isPlotLines_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+setappdata(hfig,'isPlotLines',get(hObject,'Value'));
+end
 
+function checkbox_isPlotFictive_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+setappdata(hfig,'isPlotFictive',get(hObject,'Value'));
 end
 
 function pushbutton_exporttoworkspace_Callback(hObject,~)
@@ -1069,12 +1119,12 @@ function QuickUpdateFish(hfig,new_i_fish,init) %#ok<INUSD>
 numcell = CONSTs{new_i_fish}.numcell;
 
 temp = CONSTs{new_i_fish}.CRAZ;
-CRAZ = zeros(numcell,size(temp,2));
-CRAZ(CIX,:) = temp;
+CellRespAvr = zeros(numcell,size(temp,2));
+CellRespAvr(CIX,:) = temp;
 
 temp = CONSTs{new_i_fish}.CRZt;
-CRZt = zeros(numcell,size(temp,2));
-CRZt(CIX,:) = temp;
+CellResp = zeros(numcell,size(temp,2));
+CellResp(CIX,:) = temp;
 
 temp = CONSTs{new_i_fish}.CInfo;
 CInfo(numcell).center = '';
@@ -1085,17 +1135,21 @@ function QuickUpdateFish(hfig,new_i_fish,init) %#ok<INUSD>
 CInfo(numcell).x_minmax = '';
 CInfo(CIX) = temp;
 
-setappdata(hfig,'CRAZ',CRAZ); % Matrix array, misc collection
-setappdata(hfig,'CRZt',CRZt); % Matrix array, misc collection
-setappdata(hfig,'CInfo',CInfo); % Matrix array, misc collection
+setappdata(hfig,'CellRespAvr',CellRespAvr); 
+setappdata(hfig,'CellResp',CellResp); 
+setappdata(hfig,'CInfo',CInfo); 
 
 UpdateFishData(hfig,new_i_fish);
 if ~exist('init','var'),
-    UpdateFishDisplay(hfig,new_i_fish);
+    UpdateFishDisplay(hfig);
 end
 end
 
-function UpdateFishData(hfig,new_i_fish)
+function UpdateFishData(hfig,new_i_fish) % loading steps shared by both quick-load and full-load
+M_fish_set = getappdata(hfig,'M_fish_set');
+fishset = M_fish_set(new_i_fish);
+setappdata(hfig,'fishset',fishset);
+
 UpdateDatamenu_Direct(hfig,1);
 
 % save ClusGroup before updating, if applicable
@@ -1122,12 +1176,12 @@ function UpdateFishData(hfig,new_i_fish)
 setappdata(hfig,'Cluster',Cluster);
 end
 
-function UpdateFishDisplay(hfig,new_i_fish)
+function UpdateFishDisplay(hfig) % loading steps that are not performed at very first initialization
 stim = getappdata(hfig,'stim');
-M_fish_set = getappdata(hfig,'M_fish_set');
+fishset = getappdata(hfig,'fishset');
 
-fishset = M_fish_set(new_i_fish);
 [~, names] = GetStimRegressor(stim,fishset);
+
 global hstimreg;
 set(hstimreg,'String',['(choose)',(names)]);
 
@@ -1148,25 +1202,57 @@ function popup_loadfullfishmenu_Callback(hObject,~)
 set(hfishnum,'Value',new_i_fish);
 end
 
-function LoadFullFish(hfig,new_i_fish,CONST)
+function LoadFullFish(hfig,new_i_fish)
 data_dir=getappdata(hfig,'data_dir');
-if ~exist('CONST','var'),
-    disp(['loading fish #' num2str(new_i_fish) '...']);
-    tic
-    load(fullfile(data_dir,['CONST_F' num2str(new_i_fish) '.mat']),'CONST');
-    toc
-end
-setappdata(hfig,'isfullfish',1);
+disp(['loading fish #' num2str(new_i_fish) '...']);
+
+%%
+tic
+fishdir = fullfile(data_dir,['CONST_F' num2str(new_i_fish) '_fast.mat']);
+load(fishdir,'const');
+load(fishdir,'dimCR');
+CellResp = zeros(dimCR);
+num = 0;
+nParts = round(dimCR(1)*dimCR(2)/42000000);
+for i = 1:nParts,
+    load(fishdir,['CRZt_' num2str(i)']);
+    eval(['len = size(CRZt_' num2str(i) ',1);']);
+    eval(['CellResp(num+1:num+len,:) = CRZt_' num2str(i) ';']);
+    eval(['CellResp(num+1:num+len,:) = CRZt_' num2str(i) ';']);
+    num = num+len;
+end
+% for i = 1:nParts,
+%     load(fishdir,['CellResp_' num2str(i)']);
+%     eval(['len = size(CellResp_' num2str(i) ',1);']);
+%     eval(['CellResp(num+1:num+len,:) = CellResp_' num2str(i) ';']);
+%     num = num+len;
+% end
+toc
+setappdata(hfig,'CellResp',CellResp);
 
-% load all fields from CONST, with names preserved
-names = fieldnames(CONST); % cell of strings
+%% load all fields from CONST, with names preserved
+names = fieldnames(const); % cell of strings
 for i = 1:length(names),
-    setappdata(hfig,names{i},eval(['CONST.',names{i}]));
-end
-setappdata(hfig,'numcell',length(CONST.CInfo));
     
+    
+    
+    % renaming exception!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
+    if strcmp(names{i},'CRAZ'),
+        setappdata(hfig,'CellRespAvr',const.CRAZ);
+    elseif strcmp(names{i},'photostate'),
+        setappdata(hfig,'stim_full',const.photostate);
+        
+        
+        
+    else
+        setappdata(hfig,names{i},eval(['const.',names{i}]));
+    end
+end
+setappdata(hfig,'numcell',length(const.CInfo));
+
 UpdateFishData(hfig,new_i_fish);
-UpdateFishDisplay(hfig,new_i_fish);
+UpdateFishDisplay(hfig);
+setappdata(hfig,'isfullfish',1);
 end
 
 function popup_datamenu_Callback(hObject,~)
@@ -1178,29 +1264,49 @@ function popup_datamenu_Callback(hObject,~)
 end
 
 function UpdateDatamenu_Direct(hfig,ID)
-CRAZ = getappdata(hfig,'CRAZ');
-CRZt = getappdata(hfig,'CRZt');
-FcAvr = getappdata(hfig,'FcAvr');
+CellResp = getappdata(hfig,'CellResp');
 Fc = getappdata(hfig,'Fc');
 datanames = getappdata(hfig,'datanames');
 tlists = getappdata(hfig,'tlists');
-photostate = getappdata(hfig,'photostate');
+stim_full = getappdata(hfig,'stim_full');
+periods = getappdata(hfig,'periods');
+shift = getappdata(hfig,'shift');
+numcell = getappdata(hfig,'numcell');
+stimtypelist = 1:length(tlists);% getappdata(hfig,'stimtypelist');
+
+fishset = getappdata(hfig,'fishset');
+
+if fishset == 1,
+    CellRespAvr = getappdata(hfig,'CellRespAvr');
+    FcAvr = getappdata(hfig,'FcAvr');
+else
+    % find averages of different stimuli, and splice together
+    % in future should allow user to chose stimtypelist to compile average
+    % of selected stimuli groups
+    CRavr = [];
+    for i = 1:stimtypelist,
+        M = circshift(CellResp(:,tlists{i}),shift,2);
+        CRavr = horzcat(CRavr,mean(reshape(M,numcell,periods(i),[]),3));
+    end
+    setappdata(hfig,'CellRespAvr',CRavr);
+end
+
 
 %% SWITCH LEFT/RIGHT DIRECTION AGAIN
-FcAvr = vertcat(FcAvr(2,:),FcAvr(1,:),FcAvr(3:end,:));
+% % FcAvr = vertcat(FcAvr(2,:),FcAvr(1,:),FcAvr(3:end,:));
 Fc = vertcat(Fc(2,:),Fc(1,:),Fc(3:end,:));
 % now 1=left, 2=right, 3=forward, 4 = raw left, 5 = raw right
 
 %%
 if ID == 1,
-    M_0 = CRAZ;
+    M_0 = CellRespAvr;
     fictive = FcAvr;
 elseif ID == 2,
-    M_0 = CRZt;
+    M_0 = CellResp;
     fictive = Fc;
 else    
     IX = tlists{ID};
-    M_0 = CRZt(:,IX);
+    M_0 = CellResp(:,IX);
     temp = Fc(:,IX);
     % normalize fictive channels, in 2 sets
     for k = 1:3,%size(F,1),
@@ -1211,8 +1317,8 @@ function UpdateDatamenu_Direct(hfig,ID)
     temp(4:5,:) = (m-min(min(m)))/(max(max(m))-min(min(m)));
     fictive = temp;
 end
-% stim from photostate
-stim = photostate(tlists{ID}); 
+% stim from stim_full
+stim = stim_full(tlists{ID}); 
 
 setappdata(hfig,'M_0_fluo',M_0);
 setappdata(hfig,'fictive',fictive);
@@ -1411,11 +1517,12 @@ function edit_t_range_Callback(hObject,~)
 end
 end
 
-function UpdateTRange(hfig,range)
+function UpdateTRange(hfig,range) % this function is incomplete... delete?
+% what about stimulus???
+
 M_0 = GetM_0(hfig);
 cIX = getappdata(hfig,'cIX');
 fictive_0 = getappdata(hfig,'fictive_0');
-% photostate_0 = getappdata(hfig,'photostate_0');
 
 % set M
 M = M_0(cIX,range);
@@ -1423,10 +1530,6 @@ function UpdateTRange(hfig,range)
 % set fictive
 fictive = fictive_0(:,range);
 setappdata(hfig,'fictive',fictive);
-%     % set photostate
-%     photostate = photostate_0(:,range);
-%     setappdata(hfig,'photostate',photostate);
-
 end
 
 %% row 2: Operations
@@ -1504,17 +1607,17 @@ function popup_ranking_Callback(hObject,~)
 if length(periods)==1,
     period = periods{1};
     [C,~] = FindCentroid(gIX,M);
-else % i_fish>=8,
+else %if i_fish==8,
     tlists = getappdata(hfig,'tlists');
-    CRZt = getappdata(hfig,'CRZt');
+    CellResp = getappdata(hfig,'CellResp');
     IX = tlists{6}; % ptomr_circ
-    M_ = CRZt(cIX,IX);
+    M_ = CellResp(cIX,IX);
     [C,~] = FindCentroid(gIX,M_);
     periods = getappdata(hfig,'periods');
     period = periods{1}+periods{2};
+% else ?
+
 end
-% C2 = zscore(C,0,2);
-% C_3D = reshape(C2,size(C2,1),period,[]);
 C_3D_0 = reshape(C,size(C,1),period,[]);
 C_3D = zscore(C_3D_0,0,2);
 
@@ -1814,9 +1917,57 @@ function PlotRegWithStimMotor(hfig)
 reg = imNormalize99(regressor);
 regbar = repmat(reg,[1,1,3]);
 
-subplot(3,1,1);image(stimbar); axis off; title('stimulus');
-subplot(3,1,2);image(regbar); axis off; title('regressor');
-subplot(3,1,3);imagesc(fictive);colormap hot; axis off; title('motor');
+h = subplot(3,1,1);image(stimbar);set(h,'box','on','xtick',[],'ytick',[]); title('stimulus');
+h = subplot(3,1,2);image(regbar); set(h,'box','on','xtick',[],'ytick',[]); title('regressor');
+subplot(3,1,3);
+
+temp = fictive;
+if 1, % plot all 5 lines
+    fc = vertcat(temp(1,:),temp(3,:),temp(2,:),temp(4,:),temp(5,:));
+    
+    imagesc(fc);colormap gray
+    set(gca,'YTick',[],'XTick',[]);
+    set(gcf,'color',[1 1 1]);
+    set(gca, 'box', 'off')
+    hold on;axis ij;
+    
+    % plot division lines
+%     for i = 0:3,
+%         y = i+0.5;
+%         plot([0.5,length(fictive)+0.5],[y,y],'w','Linewidth',0.5);
+%     end
+%     % labels
+%     names = {'Left','Right','Forward','Raw L','Raw R'};
+%     x = -s2*0.05;
+%     for i = 1:5,
+%         y = i;
+%         text(x,y,names{i},'Fontsize',7);
+%     end
+    
+else % only plot top 3 lines
+    fc = vertcat(temp(1,:),temp(3,:),temp(2,:));
+    imagesc(fc);colormap gray
+    set(gca,'YTick',[],'XTick',[]);
+    set(gcf,'color',[1 1 1]);
+    set(gca, 'box', 'off')
+    hold on;axis ij;
+    
+    % plot division lines
+    for i = 0:2,
+        y = i+0.5;
+        plot([0.5,length(fictive)+0.5],[y,y],'w','Linewidth',0.5);
+    end
+    % labels
+    names = {'Left','Forward','Right'};
+    x = -s2*0.07;
+    for i = 1:3,
+        y = i;
+        text(x,y,names{i},'Fontsize',10);
+    end
+    
+end
+
+imagesc(fictive);colormap hot; axis off; title('motor');
 end
 
 function out=imNormalize99(im)
@@ -1881,9 +2032,7 @@ function checkbox_flipstim_Callback(hObject,~)
 stim = getappdata(hfig,'stim');
     
 if regchoice{1}==1, % stim Regressor
-    i_fish = getappdata(hfig,'i_fish');
-    M_fish_set = getappdata(hfig,'M_fish_set');
-    fishset = M_fish_set(i_fish);
+    fishset = getappdata(hfig,'fishset');
     [regressors, names] = GetStimRegressor(stim,fishset);    
     regressor = regressors(regchoice{2}).im;
     
@@ -1928,8 +2077,8 @@ function checkbox_flipstim_Callback(hObject,~)
         fliptrans = GetPhotoTrans(flipstim);
         
         % transition e.g.:
-        % photostate: 2     3     1     3     3     0     0     1     1     0     3     2     0     2     2     1
-        % phototrans: 6    11    13     7    15    12     0     1     5     4     3    14     8     2    10     9
+        % stimulus    : 2     3     1     3     3     0     0     1     1     0     3     2     0     2     2     1
+        % transitionID: 6    11    13     7    15    12     0     1     5     4     3    14     8     2    10     9
         
         IX = [];
         targetstates = unique(fliptrans,'stable');
@@ -2129,11 +2278,11 @@ function pushbutton_IterCentroidRegression_Callback(hObject,~)
 
 %% row 3: (TBD)
 
-function checkbox_dataFR_Callback(hObject,~)
-hfig = getParentFigure(hObject);
-SetDataFR(hfig,get(hObject,'Value'));
-RefreshFigure(hfig);
-end
+% function checkbox_dataFR_Callback(hObject,~) % obsolete !!!
+% hfig = getParentFigure(hObject);
+% SetDataFR(hfig,get(hObject,'Value'));
+% RefreshFigure(hfig);
+% end
 
 %% ----- tab four ----- (Clustering etc.)
 
@@ -2728,6 +2877,7 @@ function pushbutton_corrplot_Callback(hObject,~)
 M = getappdata(hfig,'M');
 [C,~] = FindCentroid(gIX,M);
 coeffs = corr(C');%corr(C(1,:)',C(2,:)')
+figure('Position',[1000,200,500,500]);
 CorrPlot(coeffs);
 end
 
@@ -2744,8 +2894,8 @@ function CorrPlot(coeffs)
 
 RGB = ImageToRGB(im,cmap,minlim,maxlim); % map image matrix to range of colormap
 
-figure('Position',[1000,200,500,500]);
 image(RGB); axis equal; axis tight;
+set(gca,'XTick',1:length(im),'YTick',1:length(im));
 
 for i = 1:size(im,2), % horizontal.. because of image axis
     for j = 1:size(im,1),
@@ -3033,7 +3183,7 @@ function SetDataFR(hfig,dataFR)
     M_0_fluo = getappdata(hfig,'M_0_fluo');
     setappdata(hfig,'M',M_0_fluo(cIX,:));
 else % Unchecked
-    M_0_reg = getappdata(hfig,'M_0_reg');
+    M_0_reg = getappdata(hfig,'M_0_reg'); % obsolete !!!
     setappdata(hfig,'M',M_0_reg(cIX,:));
 end
 global hdataFR;
@@ -3201,17 +3351,19 @@ function UpdateClusID(hfig,clusID)
             figure('Position',[100 400 1300 400]);
             subplot(1,3,1);
             CORR = corr(C');
-            imagesc(CORR);axis equal;axis tight
+            CorrPlot(CORR);
             
             subplot(1,3,2);
             dendrogram(tree,numU,'orientation','right','reorder',leafOrder);
             set(gca,'YDir','reverse');
-            colormap(jet);
+            set(gca,'XTick',[]);
+%             colormap(jet);
 
             subplot(1,3,3);
             C2 = C(leafOrder,:);
             CORR2 = corr(C2');
-            imagesc(CORR2);axis equal;axis tight
+            CorrPlot(CORR2);
+%             imagesc(CORR2);axis equal;axis tight
     end
     % sort for uniform colorscale
     temp = zeros(size(gIX));
@@ -3331,7 +3483,7 @@ function UpdateIndices(hfig,cIX,gIX,numK)
     setappdata(hfig,'cIX',cIX);
     setappdata(hfig,'gIX',gIX);
     if exist('numK','var'),
-        setappdata(hfig,'numK',numK);
+        setappdata(hfig,'numK',double(numK));
     end
 else
     errordlg('dataset too large!')
@@ -3368,8 +3520,11 @@ function RefreshFigure(hfig)
 clrmap = getappdata(hfig,'clrmap');
 rankscore = getappdata(hfig,'rankscore');
 rankID = getappdata(hfig,'rankID');
-
 iswrite = (rankID>=2);
+isPlotLines = 0; %getappdata(hfig,'isPlotLines');
+isPlotFictive = 1; %getappdata(hfig,'isPlotFictive');
+
+isPopout = 0;
 
 if isempty(cIX),
     errordlg('empty set! go back!');
@@ -3384,11 +3539,21 @@ function RefreshFigure(hfig)
 figure(hfig);
 % left subplot
 h1 = axes('Position',[0.05, 0.04, 0.55, 0.83]);
+% if isCentroid,
+%     [C,~] = FindCentroid(gIX,M);
+%     DrawClusters(h1,C,unique(gIX),dataFR,numK,stim,fictive,clrmap,rankscore,iswrite);
+% else
+%     DrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,iswrite);
+% end
 if isCentroid,
     [C,~] = FindCentroid(gIX,M);
-    DrawClusters(h1,C,unique(gIX),dataFR,numK,stim,fictive,clrmap,rankscore,iswrite);
+    DrawClusters(h1,C,unique(gIX),dataFR,numK,stim,fictive,clrmap,rankscore,...
+        iswrite,isPopout,isPlotLines,isPlotFictive);
+%     DrawClusters(h1,C,unique(gIX),dataFR,numK,stim,fictive,clrmap,rankscore,iswrite,ispopout);
 else
-    DrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,iswrite);
+    DrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,...
+        iswrite,isPopout,isPlotLines,isPlotFictive);
+%     DrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,iswrite,ispopout);
 end
 
 % right subplot
diff --git a/GUIpreload_step1_cleanup.m b/GUIpreload_step1_cleanup.m
new file mode 100644
index 0000000..f87c089
--- /dev/null
+++ b/GUIpreload_step1_cleanup.m
@@ -0,0 +1,268 @@
+%%% Loading - Step 1. 
+% Summary:
+% load data from 'cell_resp.stackf' etc, and saved into
+% 'FishX_direct_load.mat' (X = number) in the same directory.
+% Stimulus parsing is done with function 'StimulusKeyPreprocessing', but
+% one is advised to also inspect the results manually.
+
+% Validation:
+% There is an option 'isDiscard50' to discard the noisier 50%, disabled now.
+% There is also a section to manually select (draw polygons) regions of
+% cells to discard that are anatomical outliers. 
+
+% Format:
+% Cell-responses are converted into z-scores and kept that way through out.
+% CInfo (cell info for the valid cells) stores the anatomical positions.
+% For the rest of parameters saved, see the end of this file. 
+
+% Next:
+% For the following Loading steps, only this 'FishX_direct_load.mat' is 
+% needed, the rest will be cleared from working memory.
+
+%% 
+clear all;close all;clc
+
+%% Set Manaully!
+
+i_fish = 10; 
+
+M_dir = {'F:\Janelia2014\Fish1_16states_30frames';
+    'F:\Janelia2014\Fish2_20140714_2_4_16states_10frames';
+    'F:\Janelia2014\Fish3_20140715_1_1_16_states_10frames';
+    'F:\Janelia2014\Fish4_20140721_1_8_16states_20frames';
+    'F:\Janelia2014\Fish5_20140722_1_2_16states_30frames';
+    'F:\Janelia2014\Fish6_20140722_1_1_3states_30,40frames';
+    'F:\Janelia2014\Fish7_20140722_2_3_3states_30,50frames';
+    'F:\Janelia2014\Fish8_20141222_2_2_7d_PT_3OMR_shock_lowcut';
+    'F:\Janelia2014\Fish9_20150120_2_1_photo_OMR_prey_blob_blue_cy74_6d_20150120_220356';
+    'F:\Janelia2014\Fish10_20150120_2_2_photo_OMR_prey_blob_blue_cy74_6d_20150120_231917'};
+
+fpsec = 1.97; % Hz
+
+%% load data
+datadir = M_dir{i_fish};
+load(fullfile(datadir,'cell_resp_dim.mat')); % 'cell_resp_dim_lowcut.mat'
+load(fullfile(datadir,'cell_info.mat'));
+cell_resp = read_LSstack_fast_float(fullfile(datadir,'cell_resp.stackf'),cell_resp_dim);
+load(fullfile(datadir,'frame_turn.mat'));
+
+% parse stimulus - should check manually to be sure!
+[stimset,stim] = StimulusKeyPreprocessing(frame_turn);
+
+%% load anatomy
+tiffname = fullfile(datadir,'ave.tif');
+info = imfinfo(tiffname,'tiff');
+nPlanes = length(info);
+s1 = info(1).Height;
+s2 = info(1).Width;
+ave_stack = zeros(s1,s2,nPlanes);
+for i=1:nPlanes,
+    ave_stack(:,:,i) = imread(fullfile(datadir,'ave.tif'),i);
+end
+
+% x-y view
+im = max(ave_stack,[],3);
+out=imNormalize99(im);
+anat_yx = repmat(out,[1 1 3]);
+
+% y-z view
+im = squeeze(max(ave_stack,[],2));
+out=imNormalize99(im);
+anat_yz = repmat(out,[1 1 3]);
+
+% x-z view
+im = squeeze(max(ave_stack,[],1));
+out=imNormalize99(im);
+out = flipud(out'); %%%% empirically necessary...
+anat_zx = repmat(out,[1 1 3]);
+
+dimv_yx = size(anat_yx);
+dimv_yz = size(anat_yz);
+dimv_zx = size(anat_zx);
+
+%% compute z-score
+cell_resp_z = zscore(cell_resp')';
+nCells = cell_resp_dim(1);
+
+%% Validating all cells
+
+isDiscard50 = false; % option to pre-screen
+if isDiscard50,
+    %% Round 1: discard 50% noisy cells based on std of zscore of baseline
+    
+    % Compute std of dimest 10% of frames for each cell.
+    prc = prctile(cell_resp_z,10,2);
+    STD_full = zeros(nCells,1);
+    for i = 1:nCells,
+        ix = find(cell_resp_z(i,:)<prc(i));
+        STD_full(i) = std(cell_resp_z(i,ix));
+    end
+    % Set threshold at 50%, i.e. discard 50% of all cells
+    thr = prctile(STD_full,50);
+    
+    temp = find(STD_full>thr);
+    [~,I] = sort(STD_full(temp));
+    cIX = temp(I);
+    
+    % visualize cells to discard
+    % gIX = (round((1:length(cIX))/1000)+1)'; % option: view sorted index
+    gIX = round(cIX/1000)+1; % option: view anatomical z index
+    M = cell_resp_ave_z(cIX,:);
+    BasicPlotMaps(cIX,gIX,M,cell_info,stim,anat_yx,anat_yz,anat_zx);
+    
+    % I_v: index of valid cells
+    I_v_Holder = ones(1,nCells);
+    I_v_Holder(cIX) = 0;
+    I_v = find(I_v_Holder);
+    
+    CRAZ = cell_resp_ave_z(I_v,:);
+    STD = STD_full(I_v);
+    nCells = length(I_v);
+    CInfo_0 = cell_info(I_v);
+    
+    %% visualize valid cells, sorted by noise
+    [~,I] = sort(STD);
+    cIX = I;
+    gIX = (round((1:length(cIX))/1000)+1)'; % sorted index
+    M = CRAZ(I,:);
+    BasicPlotMaps(cIX,gIX,M,CInfo_0,stim,anat_yx,anat_yz,anat_zx);
+
+else    
+    %% Round 1 alternative: full set
+    I_v_Holder = ones(1,nCells);
+    I_v = (1:nCells)'; % valid Index. here not actually screened, I_v is fullset
+
+    cIX = I_v; % 'cell-Index'
+    gIX = (round((1:length(cIX))/1000)+1)'; % sorted 'group-Index'
+    M = cell_resp_z(cIX,1:1000);
+    CInfo_0 = cell_info;
+    BasicPlotMaps(cIX,gIX,M,CInfo_0,stim,anat_yx,anat_yz,anat_zx);
+end
+
+%% %%%% Round 2: delete anatomically out-of-bound cells by hand
+% only execute once (manual choice 1):
+cHolder_Anat = []; % collect cIX of all out-of-bound cells
+
+
+% draw cells first
+% (choose start and stop index to draw; low numbers ~ ventral)
+
+%% (optional) step 1: set limit to only plot ventral layer, helps to find outliers within that layer
+I_start = 1;
+I_stop = round(length(I_v)/10);
+
+% plot those cells
+cIX = I_start:I_stop;
+gIX = round(cIX/1000)+1;
+figure('Position',[100 0 1300 900]);
+numK = round(cIX(end)/1000)+1;
+[tot_image, dim_totimage] = DrawClustersOnMap_LSh(CInfo_0,cIX,gIX,numK,anat_yx,anat_yz,anat_zx,'hsv','full');
+
+% Here manually draw polygons around cells to discard. 
+% Double click to connect last vertex to first vertex, then double click again within polygon to fix.
+% then click again to start drawing the next one
+% and when finished, break loop with Ctrl+C...
+MaskArray = zeros(dim_totimage(1), dim_totimage(2));
+k_zres = 20;
+% draw polygon around outliers on Y-X view
+for i = 1:100, % NOMINAL LOOP, break manually (with Ctrl+C, somehow the design didn't work)
+    h_poly_yx = impoly;
+    wait(h_poly_yx); % double click to finalize position!
+    % update finalized polygon in bright color
+    setColor(h_poly_yx,[0 1 1]);
+
+    IJs = reshape([CInfo_0(I_start:I_stop).center],2,[])';    
+    A = sub2ind(dimv_yx(1:2),IJs(:,1),IJs(:,2));
+    MaskArray = createMask(h_poly_yx);
+    MaskArray(1:dimv_zx*k_zres+10,:) = [];
+    B = find(MaskArray); % find indices of pixels within ROI
+    cIX2 = find(ismember(A,B));
+    cHolder_Anat = union(cHolder_Anat,cIX2);
+    w = waitforbuttonpress;
+    if w == 1,
+        break; % this doesn't work ?!
+    end
+end
+
+%% step 2: plot all cells (including the ones in the ventral layer, don't 
+I_start = 1;
+I_stop = length(I_v); 
+
+% ...here until the end of the cell is the exact dupliate of the last cell...
+cIX = I_start:I_stop;
+gIX = round(cIX/1000)+1;
+figure('Position',[100 0 1300 900]);
+numK = round(cIX(end)/1000)+1;
+[tot_image, dim_totimage] = DrawClustersOnMap_LSh(CInfo_0,cIX,gIX,numK,anat_yx,anat_yz,anat_zx,'hsv','full');
+
+% Here manually draw polygons around cells to discard. 
+% Double click to connect last vertex to first vertex, then double click again within polygon to fix.
+% then click again to start drawing the next one
+% and when finished, break loop with Ctrl+C...
+MaskArray = zeros(dim_totimage(1), dim_totimage(2));
+k_zres = 20;
+% draw polygon around outliers on Y-X view
+for i = 1:100, % NOMINAL LOOP, break manually (with Ctrl+C, somehow the design didn't work)
+    h_poly_yx = impoly;
+    wait(h_poly_yx); % double click to finalize position!
+    % update finalized polygon in bright color
+    setColor(h_poly_yx,[0 1 1]);
+
+    IJs = reshape([CInfo_0(I_start:I_stop).center],2,[])';    
+    A = sub2ind(dimv_yx(1:2),IJs(:,1),IJs(:,2));
+    MaskArray = createMask(h_poly_yx);
+    MaskArray(1:dimv_zx*k_zres+10,:) = [];
+    B = find(MaskArray); % find indices of pixels within ROI
+    cIX2 = find(ismember(A,B));
+    cHolder_Anat = union(cHolder_Anat,cIX2);
+    w = waitforbuttonpress;
+    if w == 1,
+        break; % this doesn't work ?!
+    end
+end
+
+%% find extra outliers on top/bottom border of image (can't always get with polygon)
+temp = [CInfo_0(:).center];
+XY = reshape(temp',[],nCells)';
+IX = find(XY(:,1)<8 | XY(:,1)> 2040);
+
+cHolder_Anat = union(cHolder_Anat,IX);
+
+%% test Plot: all antomy outliers
+cIX = cHolder_Anat;
+gIX = (1:length(cIX))';
+figure;
+DrawClustersOnMap_LSh(CInfo_0,cIX,gIX,numK,anat_yx,anat_yz,anat_zx,'hsv','full');
+
+%% Clean up anatomy outliers
+cIX_Invalid_Anat = I_v(cHolder_Anat); % convert back to index for original full set
+I_v_Holder(cIX_Invalid_Anat) = 0;
+I_v2 = find(I_v_Holder);
+
+CR_raw = cell_resp_z(I_v2,:);
+nCells = length(I_v2);
+CInfo = cell_info(I_v2);
+
+%% Plot all valid cells to save
+cIX = (1:nCells)';
+gIX = round(cIX/1000)+1;
+M = CR_raw(cIX,1:1000);
+figure;
+DrawClustersOnMap_LSh(CInfo,cIX,gIX,numK,anat_yx,anat_yz,anat_zx,'hsv','full');
+
+%% Save mat files
+CInfo_full = cell_info; % save in next round! not saved in current mats!!!!!!
+
+temp = fullfile(datadir,['Fish' num2str(i_fish) '_full_extrainfo.mat']);
+save(temp,'cHolder_Anat','cIX_Invalid_Anat','I_v','I_v2','CInfo_full','-v7.3'); % CInfo_full not saved in mats before 4/24/15!! %% 'STD_full' saved before
+
+temp = fullfile(datadir,['Fish' num2str(i_fish) '_direct_load.mat']);
+varList = {'CR_raw','nCells','CInfo','anat_yx','anat_yz','anat_zx','ave_stack','fpsec','frame_turn'}; % used to have 'periods' handcoded
+save(temp,varList{:},'-v7.3');
+
+%% (if needed)
+% save(temp,'somethingelse','-append');
+
+%% next run step 2: detrend
+
+
diff --git a/GUIpreload_step2_detrend.m b/GUIpreload_step2_detrend.m
new file mode 100644
index 0000000..e17bdc8
--- /dev/null
+++ b/GUIpreload_step2_detrend.m
@@ -0,0 +1,66 @@
+clear all;close all;clc
+
+%% manually set directory, number of cores, range of fish (ID) to process
+
+M_dir = {'F:\Janelia2014\Fish1_16states_30frames';
+    'F:\Janelia2014\Fish2_20140714_2_4_16states_10frames';
+    'F:\Janelia2014\Fish3_20140715_1_1_16_states_10frames';
+    'F:\Janelia2014\Fish4_20140721_1_8_16states_20frames';
+    'F:\Janelia2014\Fish5_20140722_1_2_16states_30frames';
+    'F:\Janelia2014\Fish6_20140722_1_1_3states_30,40frames';
+    'F:\Janelia2014\Fish7_20140722_2_3_3states_30,50frames';
+    'F:\Janelia2014\Fish8_20141222_2_2_7d_PT_3OMR_shock_lowcut';
+    'F:\Janelia2014\Fish9_20150120_2_1_photo_OMR_prey_blob_blue_cy74_6d_20150120_220356';
+    'F:\Janelia2014\Fish10_20150120_2_2_photo_OMR_prey_blob_blue_cy74_6d_20150120_231917'};
+
+poolobj=parpool(4);
+
+range_fish = 10; %1:8
+
+%%
+for i_fish = range_fish,
+    disp(['i_fish = ', num2str(i_fish)]);
+    
+    datadir = M_dir{i_fish};
+    file = fullfile(datadir,['Fish' num2str(i_fish) '_direct_load.mat']);
+    load(file,'CR_raw');
+    %     varList = {'CR_raw','nCells','CInfo','anat_yx','anat_yz','anat_zx','ave_stack','fpsec','frame_turn','periods'};
+    
+    %% detrend
+    CR_dtr = zeros(size(CR_raw));
+    tmax=size(CR_raw,2);
+    nCells=size(CR_raw,1);
+    
+    tic
+    parfor i=1:nCells,
+        cr = CR_raw(i,:);
+        crd = 0*cr;
+        for j=1:100:tmax,
+            if j<=150,
+                tlim1 = 1;
+                tlim2 = 300;
+            elseif j>tmax-150,
+                tlim1 = tmax-300;
+                tlim2 = tmax;
+            else
+                tlim1 = j-150;
+                tlim2 = j+150;
+            end
+            crr = cr(tlim1:tlim2);
+            crd(max(1,j-50):min(tmax,j+50)) = prctile(crr,15);
+        end
+        if mod(i,100)==0,
+            disp(num2str(i));
+        end
+        CR_dtr(i,:) = zscore(cr-crd);
+    end
+    toc
+    
+    CR_dtr = single(CR_dtr);
+    
+    %% save into .mat
+    save(file,'CR_dtr','-append');
+end
+
+%%
+delete(poolobj);
diff --git a/GUIpreload_step3.5_PoolClustersFromAllFish.m b/GUIpreload_step3.5_PoolClustersFromAllFish.m
new file mode 100644
index 0000000..6ddcbc9
--- /dev/null
+++ b/GUIpreload_step3.5_PoolClustersFromAllFish.m
@@ -0,0 +1,113 @@
+% Pool from all fish
+clear all;close all;clc
+
+data_dir = 'F:\Janelia2014';
+
+global VAR;
+load('VAR_current.mat','VAR');
+%%
+numfish = 8;
+CONSTs = cell(1,numfish);
+for i_fish = 1:numfish,
+    disp(['i_fish = ' num2str(i_fish)]);
+    
+    % for new partitioned data, consistant with GUI_050415, but not tested
+    fishdir = fullfile(data_dir,['CONST_F' num2str(new_i_fish) '_fast.mat']);
+    load(fishdir,'const');
+    load(fishdir,'dimCR');
+    CellResp = zeros(dimCR);
+    num = 0;
+    nParts = round(dimCR(1)*dimCR(2)/42000000);
+    for i = 1:nParts,
+        load(fishdir,['CRZt_' num2str(i)']);
+        eval(['len = size(CRZt_' num2str(i) ',1);']);
+        eval(['CellResp(num+1:num+len,:) = CRZt_' num2str(i) ';']);
+        eval(['CellResp(num+1:num+len,:) = CRZt_' num2str(i) ';']);
+        num = num+len;
+    end
+    names = fieldnames(const); % cell of strings
+    for i = 1:length(names),
+        % renaming exception!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
+%         if strcmp(names{i},'CRAZ'),
+%             setappdata(hfig,'CellRespAvr',const.CRAZ);
+%         elseif strcmp(names{i},'photostate'),
+%             setappdata(hfig,'stim_full',const.photostate);
+        else
+            setappdata(hfig,names{i},eval(['const.',names{i}]));
+        end
+    end
+    % add one variable
+    CONSTs{i_fish}.numcell = length(const.CInfo);
+    
+%%
+    % load selected clusters
+    CONSTs{i_fish}.clus = [];
+    CIX = [];
+    for i = 1, % just one for now...
+        if length(VAR(i_fish).ClusGroup{1})-i+1>0,
+            % copy from VAR
+            clus = VAR(i_fish).ClusGroup{1}(i);
+            CONSTs{i_fish}.clus(i).cIX = clus.cIX;
+            CONSTs{i_fish}.clus(i).gIX = clus.gIX;
+            CONSTs{i_fish}.clus(i).name = clus.name;
+            CONSTs{i_fish}.clus(i).numK = clus.numK;
+            CIX = [CIX; clus.cIX];
+        end
+    end
+    % copy from CONST
+    CIX = unique(CIX);
+    CONSTs{i_fish}.CIX = CIX;
+    CONSTs{i_fish}.CInfo = const.CInfo(CIX); % CIF
+    CONSTs{i_fish}.CRAZ = const.CRAZ(CIX,:);
+    CONSTs{i_fish}.CRZt = CellResp(CIX,:);
+    
+    %% OLD FORMAT without partitioning
+%     % load general info
+%     load(['CONST_F' num2str(i_fish) '.mat'],'CONST');
+%     % CONST: {'ave_stack','anat_yx','anat_yz','CInfo','periods','shift',...
+%     %     'CRAZ','CRZt','dshift_fc','FcAvr','Fc','photostate','tlists','datanames'};
+% %%
+%     % load fields from CONST, with names preserved
+%     names = {'anat_yx','anat_yz','anat_zx','periods','shift','dshift','FcAvr','Fc','photostate','tlists','datanames'};
+% 
+%     for i = 1:length(names),
+%         eval(['CONSTs{i_fish}.',names{i},' = CONST.',names{i},';']);
+%     end
+% 
+%     % add one variable
+%     CONSTs{i_fish}.numcell = length(CONST.CInfo);
+% %%
+%     % load selected clusters
+%     CONSTs{i_fish}.clus = [];
+%     CIX = [];
+%     for i = 1, % just one for now...
+%         if length(VAR(i_fish).ClusGroup{1})-i+1>0,
+%             % copy from VAR
+%             clus = VAR(i_fish).ClusGroup{1}(i);
+%             CONSTs{i_fish}.clus(i).cIX = clus.cIX;
+%             CONSTs{i_fish}.clus(i).gIX = clus.gIX;
+%             CONSTs{i_fish}.clus(i).name = clus.name;
+%             CONSTs{i_fish}.clus(i).numK = clus.numK;
+%             CIX = [CIX; clus.cIX];
+%         end
+%     end
+%     % copy from CONST
+%     CIX = unique(CIX);
+%     CONSTs{i_fish}.CIX = CIX;
+%     CONSTs{i_fish}.CInfo = CONST.CInfo(CIX); % CIF
+%     CONSTs{i_fish}.CRAZ = CONST.CRAZ(CIX,:);
+%     CONSTs{i_fish}.CRZt = CONST.CRZt(CIX,:);
+end
+%%
+clear CONST;
+temp = whos('CONSTs');
+if [temp.bytes]>2*10^9,
+    save('CONSTs_current.mat','CONSTs','-v7.3');
+else
+    save('CONSTs_current.mat','CONSTs');
+end
+
+%% optional
+timestamp  = datestr(now,'mmddyy_HHMM');
+currentfolder = pwd;
+save([currentfolder '\arc mat\CONSTs_' timestamp '.mat'],'CONSTs');
diff --git a/GUIpreload_step3_align.m b/GUIpreload_step3_align.m
new file mode 100644
index 0000000..864e55d
--- /dev/null
+++ b/GUIpreload_step3_align.m
@@ -0,0 +1,298 @@
+%%%%%%%%%%%%%%%
+%%
+% clear all;close all;
+% varList = {'CR_dtr','nCells','CInfo','anat_yx','anat_yz','anat_zx','ave_stack','fpsec','frame_turn','periods'};
+
+% % or
+varList = {'nCells','CInfo','anat_yx','anat_yz','anat_zx','ave_stack','fpsec','frame_turn'};
+
+%%
+M_dir = {'F:\Janelia2014\Fish1_16states_30frames';
+    'F:\Janelia2014\Fish2_20140714_2_4_16states_10frames';
+    'F:\Janelia2014\Fish3_20140715_1_1_16_states_10frames';
+    'F:\Janelia2014\Fish4_20140721_1_8_16states_20frames';
+    'F:\Janelia2014\Fish5_20140722_1_2_16states_30frames';
+    'F:\Janelia2014\Fish6_20140722_1_1_3states_30,40frames';
+    'F:\Janelia2014\Fish7_20140722_2_3_3states_30,50frames';
+    'F:\Janelia2014\Fish8_20141222_2_2_7d_PT_3OMR_shock_lowcut';
+    'F:\Janelia2014\Fish9_20150120_2_1_photo_OMR_prey_blob_blue_cy74_6d_20150120_220356';
+    'F:\Janelia2014\Fish10_20150120_2_2_photo_OMR_prey_blob_blue_cy74_6d_20150120_231917'};
+
+M_stimset = [1, 1, 1, 1, 1, 1, 1, 2, 2, 2];
+
+save_dir = 'F:\Janelia2014';
+
+%% MANUAL
+for i_fish = 10, %:8,    
+    disp(num2str(i_fish));
+    %% load data
+    datadir = M_dir{i_fish};
+    load(fullfile(datadir,['Fish' num2str(i_fish) '_direct_load.mat']),varList{:});
+    
+    %% ONLY RUN ONCE!!!!!!!!!
+    % add 1 frame at start:
+    CR_dtr = horzcat(CR_dtr(:,1),CR_dtr);
+
+    frame_turn = vertcat(frame_turn(1,:),frame_turn);
+    
+    % correction of an error in Fish #1
+    if i_fish == 1,
+        CR_dtr = horzcat(CR_dtr(:,1:20),CR_dtr);
+    end
+    
+    %% index processing
+    if M_stimset(i_fish)==1, % fish 1-7
+        M_period = {480,160,160,320,480,280,300}; %,{120,150,360}};
+        periods = M_period{i_fish};
+        period = periods;
+        
+        nrep = floor(size(CR_dtr,2)/period)-1;
+        dshift = 2;
+        shift = period+dshift;
+        
+        IX_all = 1+shift:period*nrep+shift;
+        CellResp = CR_dtr(:,IX_all); % old name: CRZt, 'Cell Responses Zscore trimmed'
+        
+        IX_avr = 1+shift:period+shift;
+        nCells = size(CR_dtr,1);
+        CRAZ = mean(reshape(CR_dtr(:,IX_all),nCells,period,[]),3); % 'Cell Response Average Zscore'
+        
+        datanames = {'rep average','all reps','rep #1','rep #2','last rep'};
+        ix_avr = 1:period;
+        ix_all = 1:period*nrep;
+        tlists = {ix_avr, ix_all};
+        for i = [1,2,nrep],
+            ix = 1+period*(i-1):period*i;
+            tlists = [tlists, ix];
+        end
+        
+        stim_full = frame_turn(17,IX_all);
+        stim_full = round(stim_full);
+        
+    else % multiple protocols
+        % process raw stimulus code
+        [stimset,stim_full] = StimulusKeyPreprocessing(frame_turn);
+        
+        % find time-lists = list of time-frames numbers for a certain stimulus
+        nTypes = length(stimset);
+        tlists_raw = cell(1,nTypes+1); % time-lists, i.e. frame indices
+        IX_all = [];
+        for i_type = 1:nTypes,
+            nSets = length(stimset(i_type).starts);
+            IX = [];
+            for i_set = 1:nSets,
+                IX = horzcat(IX,stimset(i_type).starts(i_set):stimset(i_type).stops(i_set));
+            end
+            tlists_raw{i_type} = IX;
+            IX_all = horzcat(IX_all,IX);
+        end
+        tlists_raw{nTypes+1} = IX_all;
+        
+        % This is the main data to store and load to GUI
+        CellResp = CR_dtr(:,IX_all); % old name: CRZt, 'Cell Responses Zscore trimmed'
+
+        %% get tlists corresponding to IX_all (tlists_raw ~ 1:totalframe#)
+        tlists = cell(1,nTypes+1); % time-lists, i.e. frame indices
+        for i_type = 1:nTypes,
+            [~,tlists{i_type}] = intersect(IX_all,tlists_raw{i_type});
+        end
+        
+        %% find average
+        CRAZ = []; % 'Cell Response Average Zscore'
+        stimtypelist = 1:nTypes;
+        numcell = size(CR_dtr,1);
+        for i = stimtypelist,
+            M = circshift(CR_dtr(:,tlists_raw{i}),shift,2);
+            CRAZ = horzcat(CRAZ,mean(reshape(M,numcell,periods(i),[]),3));
+        end
+        
+        %% variables to save later in struct
+        periods = [stimset.period];
+        datanames = {'rep average','all reps',stimset.name};                
+        
+        dshift = 2;
+        shift = -dshift; % circshift fluo left by 2 frames      
+        
+    end
+    
+    %% prepare fictive data
+    rows = [7,8,9,13,14];
+    F = frame_turn(:,rows)';
+    
+    F(2,:) = -F(2,:);
+    Fc = F(:,IX_all);
+        
+    % find averages
+    if M_stimset(i_fish)==1, % fish 1-7
+        FcAvr = mean(reshape(Fc,length(rows),period,[]),3);
+    else
+        FcAvr = []; % 'Cell Response Average Zscore'
+        stimtypelist = 1:nTypes;
+        numcell = size(CR_dtr,1);
+        for i = stimtypelist,
+            avr = mean(reshape(F(:,tlists_raw{i}),length(rows),periods(i),[]),3);
+            FcAvr = horzcat(FcAvr,avr);
+        end
+    end
+    
+    % normalizations
+    for i = 1:3,
+        m = FcAvr(i,:);
+        FcAvr(i,:) = (m-min(m))/(max(m)-min(m));
+        m = Fc(i,:);
+        Fc(i,:) = (m-min(m))/(max(m)-min(m));
+    end
+    m = FcAvr(4:5,:);
+    FcAvr(4:5,:) =  (m-min(min(m)))/(max(max(m))-min(min(m)));
+    m = Fc(4:5,:);
+    Fc(4:5,:) =  (m-min(min(m)))/(max(max(m))-min(min(m)));
+    
+    
+    %% compile CONST
+    CONST = [];
+    names = {'ave_stack','anat_yx','anat_yz','anat_zx','CInfo','periods','shift','CellResp','CRAZ','dshift','Fc','FcAvr','stim_full','stimset','tlists','datanames'};
+    for i = 1:length(names), % use loop to save variables into fields of CONST
+        eval(['CONST.',names{i},'=', names{i},';']);
+    end
+    
+    %% and save
+    %%% old method:
+    %     temp = whos('CONST');
+    %     if [temp.bytes]<2*10^9,
+    %         save(['CONST_F' num2str(i_fish) '.mat'],'CONST');
+    %         beep;
+    %     else
+    %         save(['CONST_F' num2str(i_fish) '.mat'],'CONST','-v7.3');
+    %         beep;
+    %     end
+    
+    %%% new method with partitioning of main data
+    newfishdir = fullfile(save_dir,['CONST_F' num2str(i_fish) '_fast.mat']);
+    const = CONST;
+    const = rmfield(const,'CellResp');
+    dimCR = size(CONST.CellResp);
+    save(newfishdir,'const','-v6');
+    save(newfishdir,'dimCR','-v6','-append');
+    
+    nParts = round(numel(CONST.CellResp)/42000000);
+    t = zeros(1,nParts);
+    ix = round(linspace(1,size(CONST.CellResp,1),nParts+1));
+    for i = 1:nParts,
+        disp(num2str(i));
+        eval(['CellResp_' num2str(i) '= CONST.CellResp(ix(i):ix(i+1),:);']);
+        save(newfishdir,['CellResp_' num2str(i)],'-v6','-append');
+    end
+
+
+    %% OR
+    
+end
+
+    %% initialize VAR (once)
+    
+
+%% Initialize VARS % outdated? Clusgroup?
+
+nCells = length(CONST.CInfo);
+
+    cIX = (1:nCells)';
+    i = 1;
+    VAR(i_fish).Class(i).name = 'all processed';
+    VAR(i_fish).Class(i).cIX = cIX;
+    VAR(i_fish).Class(i).gIX = ones(length(cIX),1);
+    VAR(i_fish).Class(i).numel = nCells;
+    VAR(i_fish).Class(i).numK = 1;
+    VAR(i_fish).Class(i).datatype = 'std';
+    
+    cIX = (1:100:nCells)';
+    VAR(i_fish).ClusGroup{1,1}.name = 'test';
+    VAR(i_fish).ClusGroup{1,1}.cIX = cIX;
+    VAR(i_fish).ClusGroup{1,1}.gIX = ones(length(cIX),1);
+    VAR(i_fish).ClusGroup{1,1}.numel = length(cIX);
+    VAR(i_fish).ClusGroup{1,1}.numK = 1;
+
+    %%
+        cIX = (1:10:nCells)';
+    VAR(i_fish).ClusGroup{1,1}(12).name = '1/10 of all';
+    VAR(i_fish).ClusGroup{1,1}(12).cIX = cIX;
+    VAR(i_fish).ClusGroup{1,1}(12).gIX = ones(length(cIX),1);
+    VAR(i_fish).ClusGroup{1,1}(12).numel = length(cIX);
+    VAR(i_fish).ClusGroup{1,1}(12).numK = 1;
+
+    %% varience/std for reps for each cell
+%     if i_fish==2 || i_fish==3 || i_fish==6,
+%         period_real = period/2;
+%     else
+%         period_real = period;
+%     end
+%     nrep_real = floor((size(CR,2)-shift)/period_real);
+%     while period_real*nrep_real+shift>size(CR,2),
+%         nrep_real = nrep_real-1;
+%     end
+%     CRZ_3D = reshape(CRZ(:,1+shift:period_real*nrep_real+shift),nCells,period_real,[]);
+%     %% updated method, weighing both std between each rep and (summed with) std of 1st half & 2nd half of experiment - 1/8/15
+%     % CRZ = CONST.M_array.CellResp;
+%     % if i_fish==2 || i_fish==3 || i_fish==6,
+%     %     period_real = CONST.M_array.period/2;
+%     % else
+%     %     period_real = CONST.M_array.period;
+%     % end
+%     % CRZ_3D = reshape(CRZ,size(CRZ,1),period_real,[]);
+%     % divide = round(size(CRZ_3D,3)/2);
+%     % CRZ_std1 = std(CRZ_3D(:,:,1:divide),0,3);
+%     % CRZ_std2 = std(CRZ_3D(:,:,divide+1:end),0,3);
+%     % temp1 = mean(CRZ_std1,2);
+%     % temp2 = mean(CRZ_std2,2);
+%     %
+%     % temp12 = horzcat(temp1,temp2);
+%     % temp = mean(temp12,2)+std(temp12,0,2);
+%     % [~,I] = sort(temp);
+%     % M = temp(I);
+%     % figure;plot(M)
+%     %
+%     % figure;imagesc(CRZ(I,:))
+%     %
+%     % nCells = size(CRZ,1);
+%     
+%     %% find low variance / stimulus-locked cells
+%     CRZ_std = std(CRZ_3D,0,3);
+%     temp = mean(CRZ_std,2);
+%     
+%     % find mean-std thres: 0.5
+%     [~,I] = sort(temp);
+%     M = temp(I);
+%     figure;plot(M)
+%     %%
+%     i_last = length(VAR(i_fish).Class);
+%     M_perc = [0.025,0.1,0.3];
+%     for j = 1:length(M_perc);
+%         thres = M(round(nCells*M_perc(j)));
+%         cIX = find(temp<thres);
+%         i = j+i_last;
+%         VAR(i_fish).Class(i).round = 0;
+%         VAR(i_fish).Class(i).name = ['perc < ' num2str(M_perc(j)*100) '%'];
+%         %     VAR(i_fish).Class(i).notes = ['perc < ' num2str(M_perc(j)*100) '%'];
+%         VAR(i_fish).Class(i).cIX = cIX;
+%         VAR(i_fish).Class(i).gIX = ones(length(cIX),1);
+%         VAR(i_fish).Class(i).numel = length(cIX);
+%         VAR(i_fish).Class(i).numK = 1;
+%         VAR(i_fish).Class(i).datatype = 'std';
+%     end
+%     
+%     %% shift CR?
+%     % shift = 161;
+%     % nrep=floor(size(CR,2)/period)-1;
+%     %
+%     % skiplist=[];
+%     % IX_rep=setdiff(1:nrep, skiplist);
+%     % IX=zeros(period*length(IX_rep),1);
+%     %
+%     % for i=1:length(IX_rep)
+%     %     IX(period*(i-1)+1:period*i)=period*(IX_rep(i)-1)+1+shift:period*(IX_rep(i))+shift;
+%     % end
+%     %
+%     % % cell_resp=cell_resp(:,1:nrep*period);
+%     % CRA=mean(reshape(CR(:,IX),[nCells,period,length(IX_rep)]),3);
+%     % CRAZ = zscore(CRA')';
+%     
+
diff --git a/GUIpreload_step4_loadGUI.m b/GUIpreload_step4_loadGUI.m
new file mode 100644
index 0000000..6ee11ac
--- /dev/null
+++ b/GUIpreload_step4_loadGUI.m
@@ -0,0 +1,30 @@
+% Instant Load!
+
+% change current folder to the folder containing this m-file
+
+clear all;close all;clc
+
+global VAR;
+load('VAR_current.mat','VAR');
+
+load('CONSTs_current.mat','CONSTs');
+
+data_dir = [pwd '\example data']; % 'F:\Janelia2014';%
+
+%% Start GUI
+
+[hfig,fcns] = GUI_FishExplorer(data_dir,CONSTs,VAR);
+
+% to push new functions to GUI, called with function 'push_cIX_gIX':
+% hfig = GUI_FishExplorer(data_dir,CONSTs,VAR,flag_script,var_script)
+
+% to get handle of local functions from GUI to use in workspace 
+% (call by using 'f.FunctionName')
+f = [];
+for i = 1:length(fcns),
+    eval(['f.' char(fcns{i}) ' = fcns{i};']);
+end
+
+%%
+% to recover after closing figure:
+% global EXPORT_autorecover;
diff --git a/PartitionForFastSaveLoad.m b/PartitionForFastSaveLoad.m
new file mode 100644
index 0000000..651fc76
--- /dev/null
+++ b/PartitionForFastSaveLoad.m
@@ -0,0 +1,85 @@
+
+%% Partition test
+nParts = 8;
+t = zeros(1,nParts);
+ix = round(linspace(1,size(CONST.CRZt,1),nParts+1));
+for i = 1:nParts,
+    tic;
+    eval(['CRZt_' num2str(i) '= CONST.CRZt(ix(i):ix(i+1),:);']);
+    save('CONST_F9_multi.mat',['CRZt_' num2str(i)'],'-v6','-append');
+    t(i) = toc;
+    i
+    t(i)
+end
+sum(t)
+
+%% reconstruct
+tic
+
+CRZt = zeros(size(CONST.CRZt));
+num = 0;
+for i = 1:8,
+    eval(['len = size(CRZt_' num2str(i) ',1);']);
+    eval(['CRZt(num+1:num+len,:) = CRZt_' num2str(i) ';']);
+    num = num+len;
+end
+
+toc
+
+%% for reals: partition and save
+tic
+const = CONST;
+const = rmfield(const,'CRZt');
+save('CONST_F9_fast.mat','const','-v6');
+
+nParts = 8;
+t = zeros(1,nParts);
+ix = round(linspace(1,size(CONST.CRZt,1),nParts+1));
+for i = 1:nParts,
+    disp(num2str(i));
+    eval(['CRZt_' num2str(i) '= CONST.CRZt(ix(i):ix(i+1),:);']);
+    save('CONST_F9_fast.mat',['CRZt_' num2str(i)'],'-v6','-append');
+end
+toc
+
+%% load
+tic
+load('CONST_F9_fast.mat','const');
+load('CONST_F9_fast.mat','dimCR');
+CRZt = zeros(dimCR);
+num = 0;
+for i = 1:8,
+    load('CONST_F9_multi.mat',['CRZt_' num2str(i)']);
+    eval(['len = size(CRZt_' num2str(i) ',1);']);
+    eval(['CRZt(num+1:num+len,:) = CRZt_' num2str(i) ';']);
+    num = num+len;
+end
+toc
+
+%% reformat old mats
+tic
+data_dir = 'F:\Janelia2014';
+for i_fish = 2:8,
+    disp(['fish ' num2str(i_fish)]);
+    fishdir = fullfile(data_dir,['CONST_F' num2str(i_fish) '.mat']);
+    newfishdir = fullfile(data_dir,['CONST_F' num2str(i_fish) '_fast.mat']);
+    load(fishdir,'CONST');
+    const = CONST;
+    const = rmfield(const,'CRZt');
+    dimCR = size(CONST.CRZt);
+    save(newfishdir,'const','-v6');
+    save(newfishdir,'dimCR','-v6','-append');
+    
+    nParts = round(numel(CONST.CRZt)/42000000);
+    t = zeros(1,nParts);
+    ix = round(linspace(1,size(CONST.CRZt,1),nParts+1));
+    for i = 1:nParts,
+        disp(num2str(i));
+        eval(['CRZt_' num2str(i) '= CONST.CRZt(ix(i):ix(i+1),:);']);
+        save(newfishdir,['CRZt_' num2str(i)],'-v6','-append');
+    end
+
+end
+toc
+
+
diff --git a/README.txt b/README.txt
new file mode 100644
index 0000000..f811499
--- /dev/null
+++ b/README.txt
@@ -0,0 +1,20 @@
+To start:
+Open "GUIpreload_step4_loadGUI.m" 
+Change Matlab's current folder to folder containing this m-file.
+Run this file and GUI should start.
+
+Main code is in "GUI_FishExplorer.m".
+
+For most fish, only data from 2 clusters are saved into "CONSTs_current.mat".
+Only full data for one example fish, Fish 6, is included. 
+To view data of full fish, under tab "General", select "6" in drop-down menu "Load full fish". Then most functions can be used. 
+
+To see some example selections of cells, go to tab "Saved Clusters", and browse drop-down menus "Class" and "Clusters".
+
+In tab "Operations", use the "Back" and "Forward" button to access the previous views. "Choose cluster range" takes inputs like "1-3,5" or "2:end". Cluster range are indicated by the numbers next to the color bar. 
+
+For descriptions of the other functions, please read the description at the top of "GUI_FishExplorer.m".
+
+Please let me know if you have any questions!
+
+- Xiuye Chen (xiuyechen@gmail.com)
diff --git a/StimulusKeyPreprocessing.m b/StimulusKeyPreprocessing.m
new file mode 100644
index 0000000..9ce02d7
--- /dev/null
+++ b/StimulusKeyPreprocessing.m
@@ -0,0 +1,401 @@
+function [stimset,stim] = StimulusKeyPreprocessing(frame_turn)
+
+isPlotting = 1; % plot parsing process for manual inspection!
+
+% Manual inputs needed for this function:
+% - protocol sequence and naming, as in 'block_raw' and 'stimset' initialization
+% - substitution array 'M_range', to substitute original stimulus code to a
+% tighter range that is better defined for plotting (both for plotting values and stimulus bar)
+
+%% Raw Stimlus Key ----------------------------------------------------------------------------- MANUAL INPUT!
+
+% Manually input protocol sequence
+% stored in 'block', based on the raw stimulus code
+%% Manual: params for Fish 8
+nBlocks = 3;
+block_raw = cell(nBlocks,1); % number of blocks
+% write out protocol sequence for each block; write stimulus sets within block in arrays within cell
+block_raw{1} = {[2,3,4,5],[12,13,12,14,12,15],99,};
+block_raw{2} = {[2,100,2,3,4,100,4,5],[2,100,2,3,4,100,4,5],[12,100,12,13,12,100,12,14,12,100,12,15],99};
+block_raw{3} = {[2,3,4,5],[12,13,12,14,12,15],99};
+
+stimset = [];
+stimset(1).name = 'PT (blue/red)';
+stimset(1).ij = [1,1; 3,1;];
+stimset(2).name = 'OMR';
+stimset(2).ij = [1,2; 3,2;];
+stimset(3).name = 'Spontaneous';
+stimset(3).ij = [1,3; 2,4;3,3];
+
+%%
+% PT/Color(Blue/Red)
+% 2,4: Red|Red
+% 3: Blue|Red
+% 5: Red|Blue
+%
+% OMR
+% 12: Red
+% 13: OMR F
+% 14: OMR R
+% 15: OMR L
+%
+% Spont
+% 99: same as black
+%
+% Dot(prey)
+% 23: No dot
+% 22: Dot
+%
+% Blob
+% 30: Red
+% 31: Blob R
+% 32: Blob L
+
+
+%% Manual: params for Fish 10
+% block_raw{1} = {[2,3,4,5],[12,13,12,14,12,15],99,[23,22],[30,31,30,32],[2,3,4,5]};
+% block_raw{2} = {[2,100,2,3,4,100,4,5],[12,100,12,13,12,100,12,14,12,100,12,15],99,[23,22],[30,31,30,32],[2,100,2,3,4,100,4,5]};
+% block_raw{3} = {[2,3,4,5],[12,13,12,14,12,15],99,[23,22],[30,31,30,32],[2,3,4,5],[12,13,12,14,12,15]};
+% 
+% stimset = [];
+% stimset(1).name = 'PT (blue/red)';
+% stimset(1).ij = [1,1; 1,6; 3,1; 3,6];
+% stimset(2).name = 'OMR';
+% stimset(2).ij = [1,2; 3,2; 3,7];
+% stimset(3).name = 'Spontaneous';
+% stimset(3).ij = [1,3; 2,3; 3,3];
+% stimset(4).name = 'Dot/prey';
+% stimset(4).ij = [1,4; 2,4; 3,4];
+% stimset(5).name = 'Blob';
+% stimset(5).ij = [1,5; 2,5; 3,5];
+
+%% Load stimulus
+st = frame_turn(:,17)'; % from raw file
+
+% cap abnormally high values in raw values
+temp = sort(st,'descend');
+thr = temp(round(length(st)/50)); % arb. thr
+st(st>thr) = thr+1; % assign cap value to be thr+1
+
+xv = 1:thr+1; % x-vector for histogram bins
+
+% Manual inspection
+% figure;
+% hist(st,xv);
+
+% (continue by default) find bins with significantly high counts, i.e.
+% whole range/set of raw stimulus keys, store in 'M_range_raw'
+% use to manually define standardized 'M_range'
+counts = hist(st,xv);
+thr_counts = round(length(st)/500);
+M_range_raw = xv(counts>thr_counts);
+disp(['M_range_raw = ' num2str(M_range_raw)]); % use this to adjust the substitution array, M_range, manually after this inspection!%%%%%%%%%%%%%%%%%%%%%%
+
+%% Standardized Stimulus Key ----------------------------------------------------------------------------- MANUAL INPUT!
+% Define normalized code, e.g. same stimulus under different raw stim keys
+% can be unified
+M_range =    [3,1,3,2, 9,10,11,12, 0,13, 3,14,15, 4,16]; % corresponding to
+% raw      = [2,3,4,5,12,13,14,15,22,23,30,31,32,99,100] % for Fish 10
+
+% 0 = all black; 1 = black/white; 2 = white/black; 3 = all white; 4 = all gray;
+% 5 = gray/black; 6 = white/gray; 7 = black/gray; 8 = gray/white.
+% OMR baseline = not moving?? grey??
+% 10 = forward grating (very slow, more for calibration)
+% 11 = rightward grating
+% 12 = leftward grating
+
+
+% half-fields can be displayed directly. other L/R directional cues can be
+% displayed by gradient. If not-directional or unspecified, will assign a
+% color-code randomly.
+
+%% correct transient frames
+% short method: st_rounded = interp1(M_range,M_range,st,'nearest');
+% but this doesn't work if a frame is different from both the one
+% before and the one after. Use loop below to force a manual 'round'
+% to the closer one of the 2 neighbors.
+temp = diff([0,st]);
+switches = find(temp);
+trans = switches(diff(switches)==1);
+
+for i = 1:length(trans),
+    j = trans(i);
+    if j>1,
+        st(j) = st(j-1) + round((st(j)-st(j-1))/abs(st(j+1)-st(j-1))); % round to nearest of st(j-1) and st(j+1)
+    end
+end
+
+% (loop until all transient ones are corrected)
+temp = diff([0,st]);
+switches = find(temp);
+trans = switches(diff(switches)==1);
+count = 0;
+while ~isempty(trans),
+    for i = 1:length(trans),
+        j = trans(i);
+        st(j) = st(j-1) + round((st(j)-st(j-1))/abs(st(j+1)-st(j-1))); % round to nearest of st(j-1) and st(j+1)
+    end
+    temp = diff([0,st]);
+    switches = find(temp);
+    trans = switches(diff(switches)==1);
+    count = count+1;
+    if count>5,
+        disp('count>5??');
+        break;
+    end
+end
+
+%% Standardize stimulus keys for both 'stim' and 'block', based on 'M_range_raw'->'M_range'
+% (tricky for cell arrays) custom method:
+% make a replacement matrix 'M_replace' to substitude raw stimulus key in 'block_raw'
+%  to normalized key, stored in new cell array 'block'
+M_replace = zeros(1,max(M_range_raw));
+for i = 1:max(M_range_raw),
+    ix = find(i==M_range_raw);
+    if ~isempty(ix),
+        M_replace(i) = M_range(ix);
+    end
+end
+
+stim = zeros(size(st));
+block = cell(nBlocks,1);
+for i = 1:length(M_range_raw),
+    stim(st==M_range_raw(i)) = M_range(i);
+    for j = 1:nBlocks,
+        block{j} = cellfun(@(x) M_replace(x),block_raw{j},'UniformOutput',0);
+    end
+end
+
+%% plot stimulus-key!
+
+
+%% Stimulus set segmentation, based on 'block' (manual input of protocol sequence)
+% reduce 'stim' to a sequence of keys without consecutive repeats
+ix_singles = find(diff([0,stim])); % index array to map back to raw indices (frame-number)
+singles = stim(ix_singles); % sequence of keys without consecutive repeats.
+% '_sg' below stands for 'singles'. Manipulations below in both raw and singles, in parallel.
+
+jump = 1; % searching for 'jumps' in 'singles'
+% Based on the known protocol sequence for each set, look for first
+% mismatch in 'singles', i.e. position where the next set begins, stored
+% as 'block_change(_sg)' and 'set_start(_sg)'.
+block_change = cell(nBlocks,1);
+block_change_sg = cell(nBlocks,1);
+for I = 1:nBlocks,
+    nSets = length(block{I});
+    block_change{I} = zeros(1,nSets);
+    block_change_sg{I} = zeros(1,nSets);
+    for J = 1:nSets,
+        jump_last = jump;
+        pattern = block{I}{J}; % protocol sequence for this set
+        
+        % this is sort of awkward. Instead of padding the 'already-found' sets with zeros,
+        % one could also search the cropped array and adjusted the index.
+        % Now 'comparison' is aligned with stim, works.
+        textile = repmat(pattern,1,ceil(length(singles)/length(pattern))); % tile, for use below
+        
+        template = zeros(size(singles));
+        ixs_fill = jump_last:length(singles);
+        template(ixs_fill) = textile(1:length(ixs_fill)); % crop the 'textile' to fill the space after jump_last
+        
+        singles_temp = singles;
+        if jump_last>1,
+            singles_temp(1:jump_last-1) = 0;
+        end
+        comparison = (singles_temp-template);
+        % find the position where the next set begins
+        jump = find((comparison),1,'first');
+        
+        %% contingency plan: if start of the new set is different from
+        % expected protocol ('block')
+        count_stuck = 0;
+        count_stuck2 = 0;
+        while jump-jump_last<length(pattern),
+            count_stuck = count_stuck+1;
+            if count_stuck>length(pattern),
+                count_stuck = 0;
+                jump_last = jump_last+1;
+                count_stuck2 = count_stuck2+1;
+                disp(['count_stuck2=' num2str(count_stuck2)]);
+                if count_stuck2>10, % arb. thres
+                    disp('count_stuck2>10');
+                    break;
+                end
+            end
+            
+            pattern_ = circshift(pattern,-count_stuck,2);
+            textile = repmat(pattern_,1,ceil(length(singles)/length(pattern_)));
+            
+            template = zeros(size(singles));
+            ixs_fill = jump_last:length(singles);
+            template(ixs_fill) = textile(1:length(ixs_fill));
+            
+            singles_temp = singles;
+            if jump_last>1,
+                singles_temp(1:jump_last-1) = 0;
+            end
+            comparison = (singles_temp-template);
+            
+            jump = find((comparison),1,'first');
+            
+        end
+        %% save
+        if isempty(jump), % reached end of experiment
+            %%
+            % NOTE END OF
+            % EXPERIMENT????????????????????????????????????????????????????????????????????????
+            break;
+        end
+        block_change_sg{I}(J) = jump;
+        block_change{I}(J) = ix_singles(jump);
+    end
+end
+
+% eliminate empty sets
+for I = 1:nBlocks,
+    nSets = length(block_change{I});
+    for J = 1:nSets,
+        if block_change{I}(J) == 0,
+            block_change{I}(J:end) = [];
+            block_change_sg{I}(J:end) = [];
+            if length(block{I})>= J,
+                block{I}(J:end) = [];
+            end
+            break;
+        end
+    end
+end
+
+% block_change is basically start of next set. shift by one set to get 'set_start'.
+set_start = cell(nBlocks,1);
+set_start_sg = cell(nBlocks,1);
+set_stop = cell(nBlocks,1);
+set_stop_sg = cell(nBlocks,1);
+for I = 1:nBlocks,
+    nSets = length(block{I});
+    set_start{I} = zeros(1,nSets);
+    set_start_sg{I} = zeros(1,nSets);
+    if I == 1,
+        set_start{I}(1) = 1;
+        set_start_sg{I}(1) = 1;
+    else
+        set_start{I}(1) = block_change{I-1}(end);
+        set_start_sg{I}(1) = block_change_sg{I-1}(end);
+    end
+    set_start{I}(2:end) = block_change{I}(1:end-1);
+    set_start_sg{I}(2:end) = block_change_sg{I}(1:end-1);
+    set_stop{I} = block_change{I}-1;
+    set_stop_sg{I} = block_change_sg{I}-1;
+end
+
+%% Visualize current segmentation of blocks
+if isPlotting,
+    figure;
+    hold on;
+    plot(stim)
+    for I = 1:nBlocks,
+        for J = 1:length(block_change{I}),
+            x = block_change{I}(J);
+            plot([x,x],[0,20],'r--')
+        end
+    end
+end
+%% Find shift-corrections for each stim-set, used for averaging later
+
+
+%% Find periods for each set-stype, and organize info by type into 'stimset'.
+for i_ss = 1:length(stimset),
+    % find empty sets (for this particular experiment) and delete ij's in 'stimset'
+    for i_SetRep = 1:size(stimset(i_ss).ij,1),
+        I = stimset(i_ss).ij(i_SetRep,1);
+        J = stimset(i_ss).ij(i_SetRep,2);
+        if I>length(block) || J>length(block{I}),
+            stimset(i_ss).ij(i_SetRep:end,:) = [];
+            break;
+        end
+    end
+    nSetReps = length(stimset(i_ss).ij);
+    
+    %% get period from first set of block
+    I = stimset(i_ss).ij(1,1);
+    J = stimset(i_ss).ij(1,2);
+    pattern = block{I}{J}; % protocol sequence (singles) in normalized code
+    stimset(i_ss).pattern = pattern;
+    
+    % find periodicity
+    if length(pattern)==1, % exception, e.g. 'Spontaneous' set, single stimlus key held for full duration of set
+        if J < length(block{I}),
+            next_start = set_start{I}(J+1);
+        elseif J == length(block{I}),
+            next_start = set_start{I+1}(1);
+        else
+            next_start = length(stim);
+        end
+        stimset(i_ss).period = next_start - set_start{I}(J);
+    else % find periodicity based on stimlus key repetition
+        for i = 1:length(pattern),
+            if length(find(pattern==pattern(i)))==1, % use a stimulus that appears only once per period
+                % find period
+                start_s = set_start_sg{I}(J) + i -1;
+                stop_s = start_s + find(singles(1,start_s+1:end)==pattern(i),1,'first');
+                stimset(i_ss).period = ix_singles(stop_s)-ix_singles(start_s);
+                break;
+            end
+        end
+    end
+    
+    %%
+    %     stimset(i_ss).period % assigned above
+    
+    stimset(i_ss).rawstarts = zeros(1,nSetReps);
+    stimset(i_ss).rawstops = zeros(1,nSetReps);
+    stimset(i_ss).starts = zeros(1,nSetReps);
+    stimset(i_ss).stops = zeros(1,nSetReps);
+    stimset(i_ss).nReps = zeros(1,nSetReps);
+    for i_SetRep = 1:nSetReps,
+        I = stimset(i_ss).ij(i_SetRep,1);
+        J = stimset(i_ss).ij(i_SetRep,2);
+        stimset(i_ss).rawstarts(i_SetRep) = set_start{I}(J);
+        stimset(i_ss).rawstops(i_SetRep) = block_change{I}(J)-1;
+        
+        if length(pattern)==1, % exception, e.g. 'Spontaneous' set, single stimlus key held for full duration of set
+            stimset(i_ss).starts(i_SetRep) = stimset(i_ss).rawstarts(i_SetRep);
+            stimset(i_ss).stops(i_SetRep) = stimset(i_ss).rawstops(i_SetRep);
+            stimset(i_ss).nReps(i_SetRep) = 1;
+        else            
+            % circshift to find intact periods
+            pattern = stimset(i_ss).pattern;
+            
+            ix_start = set_start_sg{I}(J);
+            ix_stop = set_stop_sg{I}(J);
+            
+            % find 'shift' (shift = 1 means starting from 2nd in 'singles')
+            segment = singles(ix_start:ix_start+length(pattern)-1);
+            shift = 0;
+            while ~isequal(segment,pattern),
+                shift = shift+1;
+                segment = singles(ix_start+shift:ix_start+length(pattern)-1+shift);
+            end
+            stimset(i_ss).starts(i_SetRep) = ix_singles(ix_start+shift);
+            % round to period within this set
+            nReps = floor((ix_stop-ix_start-shift+1)/length(pattern));
+            stimset(i_ss).nReps(i_SetRep) = nReps;
+            
+            ix = stimset(i_ss).starts(i_SetRep) + nReps*stimset(i_ss).period - 1;
+            if ix>stimset(i_ss).rawstops(i_SetRep),
+                nReps = nReps-1;
+            end
+            stimset(i_ss).stops(i_SetRep) = stimset(i_ss).starts(i_SetRep) + nReps*stimset(i_ss).period - 1;
+        end
+    end
+end
+
+% discard very first period because of potential artifacts at beginning of
+% entire experiment
+if stimset(1).nReps(1)>1,
+    stimset(1).nReps(1) = stimset(1).nReps(1) - 1;
+    stimset(1).starts(1) = stimset(1).starts(1) + stimset(1).period;
+end
+
+
+end
\ No newline at end of file
diff --git a/arrows.jpg b/arrows.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6d1147b6ebf57757aaa9a7a7a99b1aa4dac768d1
GIT binary patch
literal 28083
zcmeFZcf8v~)<5if(<|w;03n3{CXppuwwwl%Wx3muZCSRdvSrz_WXYCoxloq{cIh?r
zvIG(!S;EpnS$gkF?<D~Ogtnn~@{aAB<c7Zcd*1(6_j9e$Xy&Uk=bV{2bEa&1anoAU
z1V81YOnv=*rnAAHY12ohedvN#)2aonJi|KQV!{YtXvjyV&NZQv!qf-;2JV~IMJ_1R
zYUKqGq*mwif|kk6&uS&8$(JE`z6CNZ+P7KevvRIBBa_QZ>f)LI{^y-DXGp^0nH(Cj
zgvzd5LGrfsT&x|AXWMetDa_n=$>>GR1<g{ql&kSGnk7Z8E@&>E*#%z!!kuVn=8SHV
zn!I=>-udc`WQd;O()8R6biQR?)?&BJu*36VtIgqX>@x$gz!u1YKrk{7b}X<W3oMoy
z1J}$Yqd{nqE{F@FBsK6k@U(d5K)D)?#{7nLzNY6P*y(h3U=U;;NHMS4Qfqv3o?4yN
z1KEs2=Bim;D%T`UozcPNGg`g2c;?K`hqhe3atYZj-znF1rTY~j3+0qtDW}$|fQK-&
z1svpEBxDzenqJ~-OQaH?&q2j<F24xc8oTuyouoCXrsTQ>-16OSNpYpNZEOd+L#iTm
zIIuwB)%@a_&3Qsj<m*ap<`P<$)Y>Abmt@<lz1VFLyA*Bn@m{JyoY@T0{p_Il-{Tn2
zi!ChsMV0#{HiJwaUkMcHw{$fRNu~U@+y?m;gn+sUe2pjhT5btYBy^s|IuAkP2(kdN
zEI_b%78_8fMbOXE4iZK8ohC?PYg-h1+|os{S<(XlTNC{xyGzdj0=CWm%R~%H`AZm@
z9@jKwNw3!U6s2CR>3mJoXSiZ$kBBb7K(e18#0&LOM&%`?x><fb#rj>X>S#`x*Q>dB
ztDIXrGn%Vvbv>IKtlL{vbU<7S8j?^eEJ09vm#&Tk2cou8_47=HoRlxrfTXakxd!5X
zE89S=^3!h5)_enzzm>0B4VOU9B4opHCMKu-(L7GzdAEzh{V5zLAqRt3Sx><bJkB=n
zaW;Hps2;*`8zU=W+-$+S7lDKHoy(2yF{_C$#F<WJa4|RmnSCM7t&j}tWLVkhiLp%z
zBqYGwP|V#dXG)mM4IVNaYRM{oFC!5tE}`WK43Lz^{XIZDg@+g%PjTUP)(yv_3U~ll
zj7^X}Z#WjU_`1*A{LmRdR0kTzEiKr`x&R!w1^_$3|LTgStVB!bWKDGQx4VF8PokY_
zfQVFwGNLM2qm6DNv73lZm1@1`e3x=Is}#~)hn_;0C^iibJWGu&6tbz!iR2zg+5^|q
zonW{4L<2m_-REit)a4V@gwlPsbSQCogp#5NAlNIiv-@Q^<4x4~P7vySIm0CiS&)dp
zJDG4<@Fa>n2oPlF5f5dRXpHH+Cp+)+p)?$gbpZ!QrhbwLfY`dl2iY=K0?mSskoPkA
zfG1S;k!~~+&*x*Fgf-(ySOgc$bOC{M^;FU?``lq8lOzfm6_nc(Rk8@sxPi;9cBP%~
zNIqP2qv2wz7Lshu&F=^9JqojMSD*Y*HiU%YG7^?7tzfaI5<kiCyN^8{4o*e^iAkFy
z$Tka<z55i+gg8p!e?cR`>z;CV(rp3E*g;aV2!l9M-UgJAN<dIG8w>Ik`E5X5v5xfU
z0{E#_opL%EE^Yw?aUcVz%WNpmd!v>tsTo15U=><uIn9OYDWog;ZnED>+Xei^3|q7Q
zq0r{aF_^-7x9*auW+a|Mf^lCx6eHYeE?p3~ro~5f1y{@<O+!ff@>wfUNFi)K7$e%O
zEVwp*_5WAE%^&IAf1-y$%CDqLiB_6Ji|H7_@m#{nbEw+Y-BPqHcoZWe5#=t0An@Bd
z=xWA5(Uoi|0ffWm6P<dCDD`v;)E)$t`DC=5L2P-Rv*o*hzoY%zQr>+ckPd|dFXCm|
zhEo()2ep&#0)i%)bbwZXAVNVL4`<b`hEUy>xR;3txe_THXd;eNSrJ8GC6(M-EkR4a
z<r*xMfye;e4XELAy48zfU_a~ggyF2(>nT9&K)qyfS01Nvg26oup2J<CUQG!AMaM}W
z9&X@{PSPkt1aLP78et-hQ#L%+AT-=#!Lxb7ihCP)o^g3_Ulo^eR|5BIxY}^ZctFPW
zysOotFqp*+oFwp23U4(?1`o41(jfDAgvIR*(Ad!_+}r>i6_X`M#vR9_GU08w6+BiY
zLV33hkK2eSLxu2!kLUt&43)>ZERo4mFrI7>`GzNmr+q{z?<wHCOjH>Uif7$Ki}6Nr
zK_+aBw~UKE!kqV!c)sDHaG!unDOa%Jv-K#H-L6=}&*F;3m2UW(xT?7dcp!q8^DZqP
zsN#Cc)nI}_yvDk0cu>a;7PQ$wfXx)?V?s5&ohO+_n8sm^>;k6q;TCSSky1Xw;dURX
z=OcC;%aWD`od;zl%?#rK8sPTfOpPF`?ie0T5T1}b%S6qDui-A`V;T`mxtn-A-UTEg
zZXklVnP4oG7iV=Mu2C^2ks`P(B{Vn>kxo(NJO>k5mV%k2LWn63kxy}il=XxgfH#Wl
z;qz$+p*1{RK$YPQq6*~5WQv5rdWl9hO|+_Bnh|IMaeD<k=OIwptK&I{aQeI`BidYe
z)#q*GTP}CX$K?xkm(S-@@=~P>=nC3=mV8ll(N(`2mlaoB^T!*C>`Kc1VxuIxGMc~L
zP!*S04Y(Oibrmgv1YWMVG<TqcSL&`R8?fSf+tsKByMVrY6?MVxAkWls7n%=Nal_{_
zyMtz?5hGn%h-oylq?Zj9^DUJO;UR=+w@Jnl_UB=oOlV;NM<OIIhZ_xRt_w(t`G~t=
ztCO-EN#iI)maCCk-Vr0~HrmBt3TbiE$-L7}qA9xCfN0X(U|dnW<aURcRFp8g{Z*zO
zC6jKNjdlUuP4LHJ(M%MGj@P2Cs8@66voU|cPq|A>tUv^c?rJDzYlcX7D->s%pd+<~
z;$?!S+-8;~3XDp*WtK0*dPSxJ7Pei8H>jvD5hfBGl>{wRlXFrzoCCU6qGXxN5Ijxk
zJ`OHq>{Qc|WTJvX+2o{N5MvaSO?jFH+~Z+WT?(awWO%}C%3P519+pk>%~I5p$)>GQ
z&Fzsio{g5Rp0bT^M)j(v$z&qUTG3<EGSz~?dvRYjP-w=y-fC8BwnJS&FU<<x0_^oB
zaiJ8olHOb+M-@=qThek0VK;m0vgmGNW-lU(3Sox4W-U*}2*T%S<V!Bn?W6I6*G2h!
zNw&}h)MDO<PpnG*mOtUse3IS@3ciL`jJU!jAIcOP65a3<m>iEqF@K;cBNFTP$MXvB
z;*$PsR&lmcsvorHg%)q^0{ZKD)#DQUeygTdT4FYUTQs^=Xa)j44R*<%Ks;aOTO}bN
z;1yib;6N!~DYbMu(D3PDsa6i47TpqSP(f0zb^$X|OA3awHPQtW!DOmdX;}-wLaH8j
z*<C?BTX#s99E3GPjzQiaWNU=uL^Tv(8c5q63$a==U-Z~RqOTPw`g@i`s9bHqWGEc6
zq}pOTVhcfTOQ0Q9!hV|tA=!ACwZMgT(iIjhNVuJDg)2719?!~Q1h=ZBn2ZpNjU}Z}
zB$P+H6cTa86-i>KL8@q^n6e9Gr4^}Tj$pi|N9>xzPBv89!(wH!U8G|^XDW^q=p1hL
zk!YT-;AR_%6==i_X+=ot0y3^Jo_4!ROhh08ELCCnCgG5M4Mz35YHZNTw9BN((s&dP
zyQ7v^AR0n}9w!pfbeXcsDJfbKJQYi(5p5^DUBH}0#ABfSj#;E=EG+oFtWu2eiXU@p
z)|d(hnw@F_hl3^Akm9a#C@Z(@amF8x$w)jdG$Vc%t;cnN#@)`|I3<pTX=?&kSue%Z
zVYi1(5YZCsEwM!^CQ?B!+YHB(aHPUQX_ir9bRt|!1gHd@$oe@FPG=Jpg>%A!JAwL>
zT?%bf-r&4&ss+nD$0_MLrFp@PFDp9CSrQo~VJJy5ku4}~E*TSq9A)(<3*}sfvOALv
zr<h8Z+bKMePav*ddzhk|g*f7oQer}iae-{gAc_nUNv6y-nE`{dbVN~Tj*F#pLW$=1
zSh`+NX)YH}n-z^lB#w_X%M4QDd6DV?#yk~~Hv~P7)KwmGRuf3e$S{Rk%7dVpg1^oq
z4tJ)B8aWRX&AMuhf=cGI@p4m1dK+1})+(zZA`8Rqx|d<QMHak-rR`-kA(en_fG(k2
zL(E=*6C7ciJ6WjbLQOQFmi)P#V2`R5CD({LQdYwyx+pB4Y!^h9a%xE%A(s5+E?`5&
zilQwFp((<f4@iVdqx5_hafN(=Sl)<{30uTo0G&(T7cUmr996NUA_X<zfo)kw0X6#w
zU!g38?Ea9gl<opbqCJrI=>e(j3M!g`6+MwqOS3eKX(<fZ&{DB#r^A{#C*wSmDU+NW
zH%BXeUsP5rF&i2V%MPCP`C|cvjwQN)Ni^vPn#ZYVHlSF!q|ILlm4Z^*?=LYWvBmR#
zJyEg*GCJDGs=iXziNdlf1ageuURRrlD4~#D^OW*Pfa)o>mf<CTCBSJ7z9>}aQkmlA
zrk%Bw`BKSK;e%xZQ8N`WU!ek8Gay?le5v9MROm`0pbNZF)~Qz25`f)$Myhuy4D$AZ
z-sBC*f#X##*RbnUzAEsom?P+_wxzbNGfK_hg53@-P|JIeTqe_~A(GV=ETrpU4;s@|
zSG`!b*D}=}VBMa>e3@3ph;q)N!xk{Km>F_7?FN*@vmvtD;6%cP`Nc-fAh{3?oCaQZ
zyP60#vqev|njxDOuh+ox&Q`G9t9p@Y32haNff`mtT8>O8oNdAFcrske+K{%Mp~INj
zW+7FkUUOrXj3Y)ngM`Hrk2h;kKO8a=@mewpD@hIs3)yawVKbRZhl?7*wbPhTMi9fu
z<iich>eB^x7~!mic228dWvd;^hwCn<jg1%D5r5oPGm0rQt)pIAh9aCB&09-dK&1}s
zX1l6|>JnyWGi6JpoUzxFdNyJ>9KKA=okOw?saUT?oVWvPH#nM9uw<|aL4gp~RNG~U
zu{c9Pn5C0F%fVT)B4#KTFcYe+Ld!ODR<tJ|T{T-pjDT7xD5g5=1{#EFZ73r`5M;J>
zMwoI2@hi9|-ZR8SD1-Z$o{df98DQ_SZX(Z8L|*m~Ov>+S<blbS50bbRa5c0*$6zM$
zM(3X)aU1Q%u?R_o;uOKe-OWbE)2wE_O-$?-xfAIAH{2alE==N<6!=2QQ(&W1RO`Ll
zoaT4#+X4<s3}ImGz~pgDXIRj1M{(R2#hFkP7a=#^8-{!FPQjSLBY%eK0^$vy2Mk`k
zE?fr10|`3D_)ilvxC@jPuq5kqF<KC$a)A#3qn&YqIKUBr0h+NMU_KNDegMiB^);D<
z+v7&uQa0sgLuF5aXmU+Pq6xO<ErgnB33EkTdBd$R?Py%C1X?L8?FzGuD%&INa6#y`
zhiPw|wA6h*zZUYl3i))w4_H@;SrWByB0%UZ8$`R|NZgW6b{qs8+!T_ElS?B7P}yr$
z;8rtI)+j8ff-TqUQ_2p}3!6o2#9c`7UddMqNVXPGbhArFV+p04OO?`<jOs^nDjF!3
z16-+Wm&%n&y{hweL+kJt6yHY@4D0R|8K{Zd1+*g5VQ|4s60VMk;ilMVxZumDoB6c6
zrDePgD&&pYVqq7R@Jm_4-S(4FGGE|hY>SgppdlfwbkA=S&iUQ>rohULe7DH{Vw&<u
zf-F$16{|$rjaI&FG1KjWKPJ~gaTcYciU7u{<xG--b4fS~hUS9Cb7ei#8&<&e46>OA
zzgm1!@q;cFmC~hhJ*%eeqFjl{Wu8|nfl{SnZ`JgQ*?{<-l<@{o6WrzQ=oyLOPS#d&
zP|4JP@WWQHW{!t~y_%xb+hJ6%ml<BnK-F5rmME2xC|!$)X(pS<d2<47&ep?$TG5}$
zvEeXpRwyRS7!{yo-l$SwrD#LMG%d?our7rtCYwcwPO6}Kgo~{<p_eM{IAXUKsH~o9
zsun#hR?5ySTPWHBF*<Iqu@FrK9ie8b-b^5Y!9QKVpTzy{W7UP(A+=~u(n=8aWvr}X
zgqdQ6EW^c~lyj*NjU=_G*CVD8Pr!nam|etBE|+z3C;}-hD;SGec+G7FI_(zdT9xo9
zVmeSXXQ(jksm0+W>tqeDA;f!q3aTTR3>iVm28^YsQEJO76>v8bb}Q7D9AztP09I5i
zYB}ljA_|;^u?Q<_2?!0^DoI3;nu4TP6TP2qk+(knkx+$#=|nAPskC!WIN~T&GcnpB
zG_$cF<)M3BKwBoMWYMd}AvIuaWy;A42Pt(#h7ku2#-W_G#^h9_<gL(gBvyb*&3L?#
zZ0qKn7nRHzgO!Y0jBi_<cF55)+I3IWO}l}@Fr^riYgd8z8{RE?XwDnvEG<NwaiCmz
zIG~y_SfMi6ia_gbBckvkRyPM+<}$`cnmtowTfzUXkT+{dK8aREh15i}!1I#TUF32)
z(&M4bQD%j6!PY=53gd!HEPGQ<CXsBGQ3NZP(}q8%*UF_jQZC5Wu#pWWG{1!~Tis5g
z<+T;-gw#qod;ISL_C#baLk(BAY;>#HE1NUy@MbW>Yd0v+S6Z-?7PR9{=C`V47x4Eg
ztqhfnRuL63*iHw_kvbwJ1=U~21+^$vu}sk>r*b|hA?Y?V)CCM%F{zfr3T;|M+Tlu4
zDO<~wh>M0V7d2DCLbOUEWYNYsB|cTvbIGku2O>fZ+-C5CGXVv-wp#AFS?rLIOn~8Q
zNYzBF1$-%XSStLNlIl{31j=dMTFSJXXs+7M@zzS%D8to&C1;Zpk*GI~i&Dzm%Bxhe
z0O~M=1JlmJTil9{(p3-a$tGO(Vvf*s(A9SVV?I^^)`yh@HB(jFBI?Afxs0Q2hdZ+2
zuz*|&;A)ZZCLu8!vqXiujfz{Fu_m-tS)EXew?OvZ0Nk8Vm7Gh7BtUb;LZ}qALz3t;
zr>e1ba7&io673+Vb}0lcGh2uCwjUKO5z^k&BV`Uz8K+uLvyu)~uvS1tQx&X+wIQ62
zuvRo<XFWxW0g@`MhzbiNMJ5|Omyvocb~)hz)w2n;Gf-Ozn^T;Q*#s{~GQmX6u=?E2
zCSwj+w6Zgxm7AO@gI=^&DOIc)YpHCNVW$9Eqf9+0Ma$`)cMe0cvaQUNfI9(GMFmD%
zm@VlmSjf0PQ^u{Hn8Rn|FtOM!@e+w9bQ@?(EP@%d@R5L3vBo&bFv3+ETrqRK#=+Gr
zqW-jIH7B@Yn{}E4=7363)jSlA1Bo`Bm~PGnX?wdIVk45>NqF2f)W*7)SS7E?S&Rpc
zSWiQUws35~J`5#lm}KU~fLJ11wF*|3?GBz!TV>9csKG+OLN@DkL5RmIA-2p^s%;`c
z1l<<ggZd)@!;-ad4Ufu)Q+%%8=nAFk$OWw4mc;lWl9Gy{fyVcDTHA^j^d!}whlYW3
zT&krY6jix!7f>(+DU?W+D+byYm2^27v|G_ql?%XVFw+vb65(aDnQW|N)Fj;nvVulL
zhgxve97Eg*s8TFC$S9*zp17md1q`(!v>6A2#+{i?;)ak@$e@<XbG0(U#V|##10}FE
z&3RlbM?fd$fFaCmw3DzFlFde(lY^q8Qe|otyeDM@Z4znH8BCSSd7%yJH5_p^g*Z*q
zKHdlhngT-C+iuxbDW@tG1g>S1Av@;;gBA=c*qY5~B`JUoy8`<`zt$@<%0Otd4u#rn
z!k&fOX<bt!rz`Db{6!hGJUB3f40E+ov^F_wU6gf5r5r6%ix%rbI%rM>(SR3<5q6>C
zadrVU3e=H6wB)Txl2wOw9uzgxMBuUuOld^PB1O8jDCNypD!|#y2I?JCp=hYZ7Kx@Y
zyS?KbaPU%sU=e4})2>izK3jBTSuhK1wldB}HX#P;HAtpW6)Qs}LQAM&kHA=55FCMP
z4l$9k&@NfF>rIu-8gahhaCob6Bbw%VBl>oT3vTYHx78IfEJ6$l8W~M=w!KglZpLG+
zB&3E)ET9>fw_ZmG(Wrv6QB$C#RP*&rkqk%?=URlFRw@Vs)xbCmG&ljtFV*~F78oTt
za}Lp+NsAb>IiZ3xnvr6XnJ9<7hQp@ESTNwi+e+NbSDn4aA;vdqJY2RXYMdgPLwZXH
zCCdJw15$lS%oc2EV%vqqRITc!YE-O7BXW@x3@;T92?po_THrN|yOU}O7`I)a$TSzS
zW*c!ZY1>F*K*gaj9j1NhblIcG8ArSxYkSd-)(mtesyT0F#0X+wam`$?)*K$hRw@=m
zi&0j(fIb)Lz!EvxoCG!k3$?w#r|qxEN?Ef+%@(<;>Y&Mx3uxJi<(ttAOt)ZQMP*Qz
z1uFx0gjL|JAp_A`Vx`(sY|-KJm6DLGOEn__<{M>4OGmQ>r<bV(YPAR}6>Eg2Qn#Ax
z&PYZCZJ8D?ZrhUE^4^qWh$3R`QaE6RmaS2D+UiSCL@<=g<5r;fv8n@U8yPO(sG>GU
z2`;fTNxK`lFiZ;^E4xz_eNgk?vB|rDsGh2-122ktl5z4zKum<`dZwr_MT?g9>MgsN
zChWEfm`FD#tf-tUC+!B>8Bt^$!5Sa;6|7}v4Nmk1;D#aTzM!|$m|CHT3)2!7LUYrO
zth1SkTFiXS5Bd_k$suuG=ydFgSPEv{9?THJRT<(5MFE|rgYubatTzB}Ab}tdsR%{9
ziHbqh5?Gt|ThJ(1tGdfJi!94I1Hw~!Cgk*WTB9W00<#7FdM4)pBwGD-M?oRzR;@`$
zz2>WGw<NJTFgzsSa|4YQaM9Q8?I7Vv3x+w!k+eAy@I-&Yte-s=YH{G#>hiE=g_;E!
zvgJV?CaiK0gT;s_=TOof%&GQLJ6m>!Ro&0Ga^aM{D8->rfy|kMerF6)YEWB_q9M#`
zE~cz{u{T>0&gLY^Qs=x{t1cK;H>&9gM*#LXaIY>&YNa0VZlQ+@i&&)v1v5~Eb$TF2
z!Kp)F+71K8q^LB_Jt+g4Y(~H=M-2<=PjM~P(ez4{U;$HeY|0fUw4B<cj9NCxQM?7?
zgHnaAIq7OJQ*<V4nwIm~bUPnV2pj1Ax<w9XKhe`aV0LskhDRb90gS#p9uNj|*a@-J
zvk#>f7-PU9Rs*eu>Qmba8g&MmhBZ>NdlZTS4k3|~SkZx&!W>0d5*6LX0t*5TqAn;L
zY-LbeO-xu~PA^>R8SQA?E`(zdw21LdnD&bpSgho*RXJ{}Wx0iizc)th7Wt>c*k3D7
zw<OgB5SY^ldh8PDz;ZF9Zu5{0s~b#_yHi4shY7{!1BQCS;WTpw6iILz82Q;VsI?Gw
z@{$y+V6J@9MV2`}#Bz*?#w-~mWdp{OKbM3;rMx*+i^SQQ+tBo07toS{C%(oN>=*|o
zby5PH!a;)qOheQxVtN<~sZb>oPD2iff}sYH2yt=4oG3x1I0T1FRmj<N*+C7elmhjh
zP*5&oPJ?>becK?jOx-Yd-!@#mYz-ow)@ca$x`V)6EKB;&@$41AW=QFIi49bSoZT!P
zEDRr<1rK*#U^rmX02YQT=}ry|XA!;)DDKd=8PDi7UxV>_wDpLFx>>qdcwm5p4pxKg
zzQDs<v2cI4n7tz7REHB7zO}4B!y8QW&-gZ&WnBuv7q?>l7knGcvM%8NE8hl-pbI!4
z)1Q^|kMeCW%esIaIp`b|GJuDc)u58jc@hOMX$A>yx3}T~j_X7*H^`Y%hS3COGpEHs
zsT{0C@<~i7W+j!*^JQNbP_x=v(NZO1R2$8NGb}1)Fud?;T(#tli(w0Ik5)34XwV)h
zS5YovZ6+!fKW!v{?<b{|Z8gXeESR%EsBq9aAZ53iqoIUIh3k-=YCsXt>ev+rga<^>
zbyW<h_Q&J96#l1~ek*+!@IRs8x4=3$UBLg>Vh;XD7x4duXM@2yI6VsgO9A{8tRu)S
z;C~~{UjpmkbOC?uWiMX$i{`8#wq;4maq$)ub+c8}T<5?-n}AxTfe9GKwhrdHfQd9}
zPA1WySup63)IzA5leGZ*I?UDcYypYbid2NvMLW=98CW>L!4oWE66B}lLdsF{c%6P8
z(TGeUQ)YS{eWH>v7rgZ<V@@`6K`%uaQ81AGsqgZSxv;tvf^iJVt5CC@0Y;Xu(GoER
z3|re`Gu=fkqi7scay%WbYsI(|x0qv2m7){rXbepInBkxgb~9oK0-cbhXDoI4I-Hc8
zXx0*H@&!~E!WlTshiWBUtgv7bi?xKT=C;`j`GJXzm;>PQ`!t!01i%0{%jW}?il%|}
zSb4m+&VdEP9^f-}l3qw|LU4$N;;IOADgpd1K^rjSC8NbFV%{kE8Q|N44Y$`?a{*(}
zSNCGTzfvsuUAc57#o95rTVy5fsPn-j1#D0&6gIGGuvlzh7CRSVDl}WyG{|s9m9kXf
z5ksY0qF^ysct_dUWbvX!uM_q>FrDhzM4S(BT|lsS28-*yf<06w!OBlhV2ej}+kx63
zSL>h`6%P!0MH>pN%m&_DE7IwcC8C)yXMh#y<~*v^tXbuRlHxGMnu}N61+YGnk3|tW
zonYO-VP&Dr9I(a%0-7}|6<RH&kgBI0)Zl8*=Md6KJ;tXJF0z>js8E2lX*H+UP65AA
zoC$*^qSmM$2sEmxDCTwt3ejY}?X9%E^>#)zq(W%GM<JGK!1(DFc?<oCOwL}<q#?G3
zCV`Et7E=|WX4U}rM9zyA#S&3=5@~Zf94|0}53Zv?@HkbcZK+Bt69*M)z_ZlORbzNq
z2)8|vs?a9D@(!`yjH?hH0xLm6Rf_?sLA`Aynn_j&TI^t5b;ZKB8ZmFPVvR)UdYqy{
zZHowoCGFmv6jo~~QQeaVe#>OKR7<BEGHKNyI12$kU81aV5`>|osQAH{BLUYkfwbX-
zlhv$D(W<XyR)8=-l@2*9ZLU{jFASC|ID>|sgYtH>&bP{;66pLDALj|U(_0wwOG0Ll
za$b#Rs-$iQHkv!$TiT)9K!@Z)fFq*`dPg3L0)r_*MGCyn>2X`^s=XO7Ad2;C3<+u!
zSPj-}HGwyvO#2Kpor@cOO0|0mT&$dpXkDRjHUs79Y|9B1WuIL{1C}JxOjXN+n+IC~
zvIK%1ZAr9X6_Q0zvt%TY;Xuu|Nuq+VJ_(K|dNoC?wqr<}bQChW*`P9v40Oy3DR@3t
z)a#lrhYG1A#8y%|pG<neD=H6*6>F))(P9P;hSOn#w52t`?ph$zr4UGwG?ElyE}PRq
zF6FQ|d8;g$lL)8SZ8{%I0v5ETq&c0X;u&kMl;j;2S3IU6R>|*6cx^U4<IRAM9&N(C
zE+FD+5h=SXQvlBMjFmF{P(zbZ2n$LjBxqO6Nl~@NanK|M>l$Jwb)QoYHy~RG%rJW?
zHlOCAl8#qerD8zn3B?uBe9c6rBhG{#!|INRkaom!F{_$%gsSmSv51DsplYhB;=mxW
z%~@!v%GG1ms>_+mU@kjRM$lG5@meCi<||RQt7ry7+k`bnx5FA;uq$x6s#-1i5)Fl{
z6h(7Tz(5gEQPEN$ig;>F!y0$E+6L|PD7r1+2ey>x1nbj!v$r9XAgHKiuuzdeB>|jh
zHN(!-{c<7HmI7Yb<HAr+B@yLi;93gfp+q9PwRai_ZCjUrrn(gZ25RSm6A`e00ysbH
zY`Iz2QYcs#r`bEsL}v=}kyu$ske)c>)4AU8wk|6rOsO@^v1qj_T7h$<>Vn&l>Ly!8
z#FL4Jvqq3=*WGowSeMO=O-vwOt4f89h!e~OxUvbl4V;Q0GpMyacVE*(S9LOI4;dCc
z;<K31cBiwYLta^L3#gFQ^-|HAQ!6ABYd0g+G;NN9g=tocmkPu>gFr5sW5sp~3X!ee
zoD>_6)>5UIC0<SeCy;|zv#tP`jfOE+6GbZ7^%w&GKUf523kfx}VF9jl#zOMN6f75X
zo8fAPq#~3^LyFW>8?~+~)^;2fuomYtYA)DAvcZ&tvL#6>Zi^U*4AwCcO{N6KUd=?_
zKvf7d9<BuBqY9i;v|Dv|1=HzRs)YAkRbthXb(mFe(y0|u+y?yHI+%=s!roF5xF3S`
zy4{+E>UxrOq#7J6Bx*KaF$C5RS*+H8lMBbPT-+fJOlx+f3_klG&7PDkQA@hBrV~_;
zY9OC?0NbA82y+^0^^7Gb0JYs7-foF`fGfjlt|DHuWU$Vb&+Y^Lv=Rv*SPnxV%vP`$
z+@2Cv;TuXTUx}ve5*?SdNTXCFqLL_3-riVGDp@6?WMGC75tUfR;uI~VT#GZPG?G;D
zCPtN*ny2XmD}FdSRVg|>u$2?sh6N^LL4<D@DX1E6L|CvBj_<YDOwJZ*cE{hH`2sEs
zyDf=g$B~efkfamjDWnNUn{i1sM=7G3D_NFpkr)gnoz<GhVsArHNX{jMC>+T3>T(Dv
zD{(tj_fpYjqCtYykJV^1X|QmT;>94C>`*I_N)kgfzL;$#1xGfNRO|jy5cH<?SeC5-
zgHaAuVokN^@3{!TZWhc{u8@-M^g3z7=Rg`Yy+!j8REh!GVgZl0V6#=hK<4*c7!^lX
z%D<MbS`X%2wgUJEi~U>-mG+e>($`We9u_7%0mBb`@lYgW$Yh(eEA4VDjP{(mFjbLh
z9`_(Y5lz%%loz5yI*?QwSYRd69D`J~MYqAaK?|6GWKsms#8}!FZZlbD9!t}4TZ2ey
zanKsOQoXfvn;l73hnY2NVwuFMz<O-O6UB1P@Sqr1D7xcbhXr(8)!$RuZjpamt0gcy
z3zafx5|j9(GYf+*F5AjOtdo|sG>Uk!ic|3l#VW0W_RQ7=v_c^rLx87Qb+97WS!NQB
z6jG&@rS4WOse+{<qSi7j7@8KYa&gyIr@Vs$eYpyI?WmdMlx6}A_sqF=GEr^nEM#@W
z@<OiFDpEwQYy%qy|06rGm$*w|_mk})<`COiL-cX7RxjC;v|uj^y4h}|eX3b&B>XH`
z+ek{le3w&nJr=H1z~r1W2Sc17!Z9%+r0Qs}H(?FcNsh}gFkJ=PxGKe}2G%C9#$fh4
zlfc%A_#&EFM?<h42Sx&K&-?Qv!sRFmO1i)dt)n|b+okY7ZEXgGsQa~4+NKst(R8Ye
zCvxG2Cz0<0<}_cYE~81*0+<Y3pe__FB`MRXMp%nzG;6M<EN*`aYI6>;ZWryU7qyjZ
zHg7k16$*(h)Gj;KjIP(!E}&lbJ6w7k@2ItCRYYf8rh&h}2>MgyRM{gbQ4Mwb6JYID
z$Q71cq$Lv2H`~EcV4t@O)(TB>bf~WO9FmaX$RTx{HiyW10YYgTn2OIgna%qChn<mI
zT=#=>G>h$mnG<EPEqbClqA=EwGZ4%L=vX~bwv<CE1y&aO%MLJ^>j6tpFk8_M)_^N*
zl!n8ZR;Xuv72^)GoP}uuEQKqkAbU&4Fc!hWSkugkA*)hxfPs#)Tx*ssO)zGv7F<TW
zlCn~4HfIO3v@zfku7Xw~fM`&!$YPu2ta{uIrIKL-%m`v&N)wV)iV)MDK`!jyZ@WOV
z)fLK5eXs+}W5bn5iO*EP5`;8g<eOl1K_C$0FniGNX*l5&nEb#p-C|eFF(K)U*)SvC
z1&r}zE?oC1I?Z={gPw4^4JHO-P67fJqE$8IIGA>BgGIDr$v{Qe<*d?ruH+XylrvT>
z+AY4At(e2BHqGxFnDmcl>zWtLf!K{-mMsMJb_RDhzxMW;{&Y9$f3hvRvl02{dwBmZ
zH)n@JzrD-4RDynKhjz6V&28Ot|BD;DtDUXG(B_@PAo;9c267u<v;A*v^xhW8ZGfHj
zTX*nxcESTLg+dE_YPAMTxtxz&0wU&1lCZ!@+N?I*ZbJ#u>bAi!fj9}5-D-8hq}_$M
zknYamZL@Bb)&MN2W$T^I{R6PVe*ilu>(5{#x&+Qw;K2f-E%5a^z(z^16WWQ`oK}J&
z2^e<5HXK8p7Sal%7R=S#Bn?RZh5SGJMl`1iQs-C$m%>-8OUk;YfnD8#xVhLrdH=Kg
zWKNfioIvSXsk_bp_cw$u>gvS;X{#u@h2p{TFdLQ==UFWl%RC;o+2%QOIqN(@$XHR-
zk%OJ~+#;}~3n<CpyjuzCNV)e~*vhYtlsn?Ua0&%uNteZqSb&!Y#=tlXhY^?CM%e7Q
zjogepDC;jFYyY)GdX)V_{$Kix{;NXi!TjX?R*20~8zhxMoB9W|)hkuO*#NCUFFLI2
z?B<W<D%-|xegV#q_>Xq`SHYZAZ2_O{DAtlH4-}-s;6ZU)Nd}+jT|ju91cUVAk5|#3
zW&L&hMj@yE)|n4IsBWrX2dQdet-<R#;O9_lzjp3I4{6(^zYdu#fMH2aSORV4)aHnv
zX8EnYVOs%i)`p)5{Qs<LkhY#Q@yo6Apak8cV+J1$v1riC1+f~h*Xm_j*F*`NZZdE*
z$#0|>V9NhlQ9z6L%S8Y$0O{6k&kzBT3jnP`J)6@9q)&r`W)vx-^LlHBbG{QC9<oiM
z-=w!!3UE{nI3)+#a!k(8YU1XK@_%~&sdoLJf^P=<-$yoc;t$>b$F4ubrp<}|$F9ws
z_(S*qvFi`9X>+3gv1>CY{?Pq@?D|7&+MMXO*fo0VL3?04uy|$z9HY1C&9O<J%XKgv
zjZnUz+tk-*GL7{HB~|O6U^11|njZBKGm@$Fj1f<m`c0!u!%eWs#AmDJP>hOq&inJZ
zT{Eg6Q763h`to^j03f)YndhZv%-9<Lhe8vxWxWQ@5TrqBs{l^Q1L0#pSZUPCo%oF)
zJS8J{!u>-#VI2?%!aH=r`CfRRZkk?rzfM>vsRBsTp}8!SgiiQA5dN!C&w=oeAP664
zNVx_GKMTUMm3m16;jck>N-4)zO{Sq^JMpz#wgAEw5FV??qb?9$WHOB$n;#U;3<}qB
zO+cr~<kHG5Fxo5BX6%{WYX&$b5SszcPf&8T+B_PZe#q+rIGeCk=GB(T)GKo*muXUm
zw=)2f?O?!apNGr`?B8<z=F2Uqeml^SZ7;=VOjl-o(;nLzw>9t9IBl89gna>cbK2Ip
z%y}l$O~;x{JG{I#ZuZ|yrtOX~nb!PkP<f_x^kvYoiBc{*zr)Zi*Y76SlJeIBgT6Pv
z^Sv#+n?d$YE1c0`YF1Og0dq5|;PA=ZjCns*;{Q71AhHJ0u@CSw<iI!^d@2F7Oj7ee
z+ykzX8It;I68<l<4WgkF?CCWKOuqa_)0Bhfn<o5gifPD4kC}!}9%CAEz{TKA-xj$A
zM<q<1n#b+-Zcpz)7`*TPZu8TBDtPR#O8FU`Xje2oBU{&vP8igQ&d)H@Xw!D4DW)Aw
zyP9U1_B72iAtr}up=m#p+vGPzOmS1nbdV`)Dws+p-PAB0X8M!qIMa!y(@aZE=bJ7u
zU2eL{wAyr?=_b={rn^lKm>xAfX?o7|ifOHBz3BtfM$?z3@A~@sM)ZyAo7}f!-)?<#
z`pkVuAJ(^|&)pa7i}t1Z4(ZGHseSdn!}|W*cVgceedqUG+ILmowSBkr-QD+a-&1`r
z^{wmM(D!NIxBWx<$M#R@-=%+#eyAVo$NK~Q@%{t*^Zk|ncK<Q`r}UrOzpVf7{cHN~
z?tirZx&F2NAM}4Q#582gkZD79A2M$UHpDd~G9*1j9HI|7V#o<Y&K|OC$m$`t4ta3M
zGeg!6`EbZLLx&HYJaop;c|#Wt^$ul+W`|aW9x?Q!q2~|1a_E|&_YHk!=(?dBhyFNh
z+_34x%)=H8^9@T5%MWV|J8sz7!&VHtaoGLCo*TA)*q6hH51%%C?r_I&@9@;{;_yR<
zpEUfU;nxhmYxq;c-x~hKh!G=p9AO@@ctm6bKcY6`xDn@#`1^=EM?5`Z{fMtejv2Yz
zNMs~6GCfiqdGyG$N3I-s$H=EgzB}@}Q4>bZ8MR<kWRx(fHR{w+myf!2)W1fpANAen
ziKFL^UOYNJx;Xkzqt6|^di4FHUmg9~m@#8!j&Y7*#z<q18guTLe~fu(%(^k(jGZ)g
z@3F422aK(cJ!R|_WA7UK(%8?&jT^Vexc$bZ$JNH2GVZEz_l|pg+}GnLkDou@H(nTj
z<oFB5uNnXJ_>Z?6wcVWU@a+!TuD#va+g-oizqb2u!l(&*OmI!$Cmb>1f(f@wcwxd9
z6DLoECq^cgCZ05L<-~_4zBg&aq&+6NCkc~|nRMBtdndg$dC25hlU<Xulm9&V^2zs4
zUO#2{l(|!UQ;JhgoO1P)C#GziI%%qHYGP`0>IGA8pSpJ1kZF5N^G#ExojUEhY0ppl
zcKcnn$G6XIfBg2Vw|{#3uXfmZhy8ZQ?Qp^l*X;1z4&UxLV@LOn@{XtPc+-xr?KE_!
zy?0`EYV36JPWSKh;m%WbUbu61=M#3mZs(V$_fOw@dUX1s(=VI;*z_-U*>xAsF6CX$
z+vVO}KHPQMuKVp;-1V$o@7Q(2Zj*OgvYWKqS-ai2+xs)7&cJ6VGtQlH-;9mBPv70U
zdwusycYk8{?`F=O$<91_<~1{4n>BhCHcOm!)~tJGZJfQ^?9lAPX0Mw4@|;m~usMY}
z=gfI<&R2WP-GkfXFMHgy$9r>knCqL{p1W%9t9y>ybKgBHdoJ7a*}aDEh3zHpb<ti=
z?%lsPx_4pk3-<n(x!-IzOXiErPwg{oALl;mKFjubVcytz_&j6Ys(I_?Pn#c{fAsvD
z=6?*$hSJdK(1XyA7Sy6xF1Ne_PlEmM(eSPCr^wz&7P$a<);i8gS&y{dZ2iP$w&iRW
z+g?N`qapM!=soBUc86WJueNV+%yt~&xX|$eHU*=xldy-J!<>ZkNat<NZx+}W)D~Q~
z;NJ`9EtD6oT=?Fixr=g(E?@N4;+c#2#g{H#yJW_aLzet)$=ZE)-<RKa*}iY?H)}s(
zzZLtvi|>UO@xS9A6Ba@zZX~{PEpQ#~x{Dk_ddL&VC)|_VtowZT>(optPpzUhdQi`y
zp1ZumyaDg&-WPnk__DsMd>j1^|Ka}o0%HTQ!1;kUgL?<{;H{zlkUw;0=+*F?uo}K8
zVgj>{XGUJ5=h7AWpUiNEW-egfiNevtq7TL<#}1CIjC~m=;-|)6VfSR~>^+I?5(gx%
zN_@qU+!@^3B$PZn`DkjVl$5$TJtCb*uSkEfzkC0s`>#I$J>a+lo<DHzfz1OSK4_<d
zl!I<Rc)Nr7gRehi*dg2@D-ZdekMPU*FEYN&g_(_6SN5Fj2f}{BS;B_g;@lazcg4lx
z8RC2S#rZSy8w&dt&Mtf?xuo-@Pm12+CB?7gh<t^-N#T@hN+U~$lx|Tcsj_;XwyW0E
zo+_Km$CcMs7F3p2HtGR=MYXSbVD;wO6ktp}R^PLJY<-=v#JIrtx{+vH-<;S~n~$~j
zZXMrxw@tN|A3F5V?4kD@w)<g!I&9tH_~Dlx(RW1Vh<lElb>uNeu0P6i)Rli4{U`ZP
zj~{J0`i!GL|1<aJTaVf0n4^w)>sZgRtBxCgoPOL3e_8mKOa40auhL(iINo~vdB^{F
zf^fnkCt6NC`^4{0%AWMd$?(bNp1kQ4@suY{wV!(NX~R!bPka9KeNVsgjEQHo&v@(1
z(3xw_nsL?%XMMi(;H3|rZ9Dstb4H(2KWFW^!E<juZ}xeoo%j9uh4Y`gfVkkA3#VUr
z{Dog#BwY0L#rs{n`jTBPIq{Nj|0ezI#Y;Vxu35Inva>H6ewlIEJC`Ree{ear{Hhf@
ztvGST4_ByHti3XN<$YH<uDWXF^p&Ts>R(k~wc+mv{Qb$R$*XT(J#Y1;*G#+S#DAFn
zQUAyL*B)~1^VbEgyXSiB`qekgx#7YaC*OF&n!YuyH5+d#+_d)Q)Xh)b;=kpdTNmDX
z<3H#B^NQPczwLtCr`~?*9i#6!=8jEww(tDnuJT<U-krbut$Pl>=aqYt_da_cbKev9
z2kw990qTK!AH*NL^P$BL-S+T;hi`cVd*r4^?T@Z`41Mgz$I-`ce8T?3ntwU|b@P+X
zCvSaf(NlLkz3<a^KSMtAz_Y$*AAK(T+*8l9&%gM>0WYk5QF!samz0-2dD(dRyH}2S
zb=a#Xyf)#rvtOV7`m#6ndgGe4_O-XKBiB9lX6(&Z-pao9!Q0ihzh8gMJLBFt``z8%
zz3M&Wz1ucW8=icB|M%B_Q2yY%50Cq3;zt*LyywR^{CmHDKejQsas4OyC!0P!>9ZX_
zyW(@(=l6U;fARX4r7yqx>V&U%{QAmoux}pxmizYI?;78a`2O4<_WI%09|J$Wx~a5j
z)25G2<GTk4n}oZ7Ezq~=71J@Mu_H&095r(6s8M4lju|~>;?(WNj@@qR4wENOojiGm
z?I%qE*QD*Y-+B73(|6u}_UyTHXU|5hRx7%2%hfk-%$RZG$4wYNe!`UT+l`+xWx~XX
z6Q=AkWy;hkQ+AoN)Ar!nY0ivUyY9N{tQmXmIcLtEdm@MhLAK;yxals_#F3^a`}&6T
z%{28->>Dz%Z_|Ci<mfZ?_YWCt;|w1$bXebzk^R8H=$JBtuD%JTz9Ien!-kI<Iby`H
z?u)+uAw!2v9KJhz`lJzf<*CW+G=~4vrPt0xW=$cY8U5&G*Ij@28P?dDHS``=wn|Rj
z{^>tIGut-tXV`$4mrY{<9sLuBOf>Clx^cfj-|YqamftRZf2ZKFueoZ~!Xuw?|9j4Z
zcU`Sr|K&esaI5DXu+p*Y`nz7nzL@<qKYP_%)OX`P`&0h;7oOUGUFw7UDd*4rMtAOX
z;lrogxVvNZKCkTc)w1jFdH*Hn9%t_GW@P5KAAdIW*!n%Uu3PXd7nu`W_37LCEv++m
z@n(-b-h4fE$YVDZ7c6-BirG`qiJpy59eVRW4t@F6saM{y=FYQL9yDz7X&Zi=<=8*<
znULD|xCebdoblzTzx|Bx0rFq4ZG7gw4Og6gEqUMlpDt#uCO=vE-okP73y(aTI^+1g
z-k<)Sc;eKf4e2%g`*+^@^J(MI4^FU;IpXv9o*$gG;j-7hy>{*o_q!kbaM@|*7baBa
zywOH4S-au2%dT2+|G8H@eZ;~KZW?mnk&m7e7<TZz+=E9i+;yd~;S@5Q_~+*`%DY+i
z{@Ti;02SwzmtAu8<74LU_WkEi`%XQXJ$;i&x%rWM|9r!X&%XNp(#6NQCqXZ8w;b6&
ztr_~2LYPboiYqr1)}`KmC%0(cmzSUZ&;wU2eX;KaeGUEEUb}e?9+Usmw=aL}{^s!m
z3s<cpE`9v_eOLcu)v7(eTKUnrZ(hfLaQ~Z2C!e`t?z!i@fn8m=?DZ3#+j-SHp<Oa>
zP2cmj_b+(j)46A@zUI7@-yGh)`f%d!*qbjGe(+xK<>Z4GUQ1uSA0TK<{utYLXD_(*
z*x<DE()<GWq#x!-o9?gW$g}H5JM_mk2nX+TO0jkRF@Kx!Q|kNfz3};Ew>)xkQD&b!
zN*{M%BLC<Og*zweD>5tATz+!(PIl$$&vr^bvFCcf65UH&6@U9UG_vUS<fOHU`m32E
z&VJ>ihb|s9_srdHzAtj0NB;Ws3(S=wJIB2J+p%kR{OX0)CR1}_?O8S!ec*c5|Io~k
z?tMEOdt^O1iwn=m7011E|H4(vBd3R7jqUc?F1NmR!g4gmvE0l@S=-V{^KW`&{h@1a
zVZ*0gyWrMOieJQTdsLbD<hAw1uYVa@P3B+Q{fn;_9{Y%T>m4in=Q($owe-d2`6phw
z=s0BN%4IV%tF9U~?y?#09J>00%a*4L2mbIy<e%zp?BVrKZ6i0Fr(FN)$IpB*?fv|-
z4?lYI#j7gotLHqYEe$Mr^qv#lGuP~Y`RZ$yS`NAzK4td1<bm&&KAr4OFG%mXbHq1h
zuXWn>(@WHii%uT4AM(|xbtmu79C6NS<IZD_Wp{k+8=LcpC#JviZ+PREk2Wr+o+ckX
z{D66{9hW@+<jk_e&fDQtII`=V_}+cHHg8L*_Vk0yt22|Ahvpe;7T@VSWxZ?Y=jfXc
zmu_cnJ!Rdbz)l<F8yBDQ&0~$3ONSR8Y6PYd3vB6?qYgUxgymN)U%CG7tCvoG>EK<L
ze>QCYRQi&2*S&V#)F-cAe*6KK_#XExd}76uUmv%}y_09=uiu$D`JLh+e|fQ^AiZmt
z@gq#H*yX@AT(fbxeE7J^XPZp$Gsga#OebCP^&@u;OU-|4)^m#k#H`!Q%n8c-^n+)f
z<e2l|%A1IP4gc)be^VO{Htq50CmOQvlv74~&pe$umHqDNy;~>k`1Ls#&bfc*#EZ<4
zjVVW*gFZs-@Zjx@<a7Gh6X6)L^VE-zd~yHxmp}N9Yu4@KhU|SACZ2F%`<3PXnIBNz
zYgat<RY>@v&dz`Rs`GsptvmgZ_1`_R;+?U_U$AoDys{Mj;mtMsK2{A~<QRQ@{EZv^
zZSAw!4}N<7#@+9_c0+UFg0(lEW552M7ngB|zw5l?&sEPohut)O3^T_a{q(YnPf{Ox
z`NVhoFF~%ny0Q0H>XT1D9=&rN{I7d2soS={Xd3<K6l~t&HAAa)=SLqLa7ppN_4beC
z^s_Rv)3eVWf9Zs&!ToRM&zCK=ryeBlUw_et+yC<1sfQbD{iWMx-gxhyKW6pk7JgNk
zNj{M~IQ8f|bJxz=UwQOwb9lFpK4<p(Jh<R7$K^lH`lNB;OFgEqdt5#}{kGfv{E5rQ
z?*|+t>(B7aYFwr-zImmFGO;NKY|nho-2TpO$?CX!0*Af-;`6I>XI%8-4Uw--`SQGP
zW8Upgo9Wb!pC&)BesN~`9XHlg?;RUH_<Z)kuG`+exraZ_jEJm=r8m5=+lBWv-n{R-
z;AKmns%$b9m;X4qJ?`cq=~r($C;89EZhY#-*QZ=@{Pr8qoA~n8r}NJ-ckjAW%&^bB
zbnM?ojFXek$`e7^|9bB35A{&D`Q2{O!AD;G>b~b+@Y(eC^v~}9>f6`u{pPAW;kzqG
z`;j*$Xo<xS&2BvK<V{1rZn2H?FIo5S^8FuQzKA@a@9_2Czmj`w;wICE&_C(pPuw7U
zk<-pxKlkJn?|r63A02!0*>i7o?r`&sAF=Z3v87W-y|e$D%X9NTLO(f?>o53M9C9YQ
z<cxynn{zIlC=Z$RO8uq-r88E(v2gF*4?gSur%zl9PL|;ATanz)G;G`(hrRT2^`r2g
zHs4=8>7|#>{4jj_!2Pr<4*x7K{p-dBBR1c4YhH|8yTCJW->>n8-MH|(*x$C?XW0$m
zyUCX~2JTxj{tw{ynfOd#pELd6EZ6VwO~2;QbJDjwdcsw=d`u}HKK$ie--@&M+I{K9
zN1hW-za{zQ>Km%uu>F>FR4w<4dehFVpWXl6!xugL>AAN~T{OP<p58p`x=-Jjy65{h
zN!Og@_kA%Z04FcJ(`Q??Xv1Ti{m^mGf5kkz|5dLcSFHIVm}wsL#goTga`xA}yyoL)
zUt@0?eZw=)tvvL&YmQ&_(GOQ2<=^)W<)TXi-xoHS-X9*h=b0lmT(a^a)2thITJqic
zCC~i1`1x+ZCsOQtE6?%Q?tORh&R_pyNNrrqnn<-X=cH|Ke{sdhcOL(cfP9uey?n~8
z=Y4R^b4yQrue{^mz80?ap8LR;w@*Ik{bk>7_m6GrO!r~L#n7G4f4tk{p>IxEJihhj
zcR#4>zJK<ObJyN{x9QyVt9N?s<9{yQer)OUz!fh<4*UGueGWX};}`!%Zn&ygd^fjv
zpUB11_QO3pnJ>KUp=1AY_Sf`R>a*``|M7|&mL30c>+;iYz2$-%E=nQmm)$>tRsS?g
zUok28?=<&Ae)T1LTqaJsb<g*%d~dH;Lhdnpt<7IQy>N2)@ab1H0Zpwdr(N;=;?hT>
zHkp{e-23%H^5%MDul?7(e)8(iotLuZeIDFcI^mAxFQqRY5qs;s^Tswd8aM8n*N=ZA
zJ=gK%yL|`ET*Xd*?(SH0#r1pdeEZ9nj9<29**>W1|MLBBSI>ER<B2c|pR(eN$fR4T
zYu2wl<neo_zcFp$S)cstFOMxf{?t#V9(73UUFD_7U%2W;FCOtp^x(%mJMXRTB|cRC
zVD@oWZ1?(o$A5ERYSonww^m=W>vb>CAMR_se$%I4ot$%zEA73GjqEdK?dhlF-z57F
z4Lq>-18bL_{o>mj{~leiul0htm%qk*BF}&P==i-%@MAV{<3}&QpdJ47rKW@9sb|dG
zA*UU1sb~I*M^;VxaOqBJYS!$`6346;KblwNrW}-WJ${^fmj~`R_Pftn<e7yB!HXaJ
ze($4BrH)>DQ^krhhrBd5m!16fv+&d-54_-%moH!O()}MCb}sU+d(MLQQ@h7czwP}O
zROg~~(?+7k-yi4yx%RB7*ZRNsaRaAE*4;Uxb@Jc3N-%!vL(ks-$X_p>I^;A{|5>Nb
zJ?Q#j-`+0Ge1F3c^M%6?xOt83t5Y5v&q2qJ`TM$w6D{Q3i%is~uM^k3*bra)`jNMv
z`2<<kUiIV7w#yD*g7~|Dn}3EJOUC2z`?pU|KMH-WW!#%O_5S@gEN8z7Om5tnnmjp>
zx_A2W*a;5~J!|#<x44LuvO6mG{SCDl!mjR-&6^(mPHP^=^M2X&N&gvA`mTStY!>+F
zS@5Q{oolbWcdxzv&iYJjg6MOpT~fC+tyZ0?)Lim&(#A8feJ)n~9_Rg8U9v@8a%P!l
zr}Vm;R$jctarb<?WZ1****;r>^iMA<dsMgRy>SxjDo@eCLSL@7e;4O%HeDLH;p(o6
z=)XtZ_oi{4n0<QB{Y@+K*WFRs?a(*rda!J^aESl+pkg(xHSt;td3QwRm7T0w-D)!B
z{FB#GE4shNuDs`<b6a05H`ID#?1ECB?!{}<j9ky{sCs*2$*sO^eY;pzES@=4V0XyG
zoicN7@mtI+d%gD9^?<2jzuGT*&AQsXu*NTMxq7FgTo-Hp^c|ayug-k`;cNEl$BM3D
zpZD2nKg~N_Iparn>ee9BD?*zdh8veJ>@!a~w&>4FnY;1f@dj?6zSeHL@#5O1$NFJs
zeUdV(?|zQB7pkRmaDHC-tRE?t{T=<aRlBBkuZ?(8WaWBs&s&)jnR?f|d!KE*I-PS`
zX6(WhyQHrDxduG>^bpsVMOs=)iyn0^73K5c=5;?B@%F35qfM8czCI5#m$#fJ0E=97
z+G>^N%2$g@L=BBH=D&4U4m=xIQr5hMkz-%My3<j8({6l^ejaPx+5Ir|*SeVKC-I80
z@kLWV9eVoXRcVCb!L5eBKI-3k@~n*Yaq+GTQ5pB6<Bby<=d4~EK5?yY*6IgN3(o%h
zweR2apv(C|(T0B~>91Fl*|e%oHs)cV1JEJj^Z8fVDQiUES>3g9eQ({7{;bN0oPwMW
zb=@PUJ#8&L81lvCqve|Nr6IGyL;UDY2l08uvd&Ji%$mB%@49cWnLR&qNxoLm&RetE
zzPwtoY}u;T3!yG2AJ%V`0jH=N3q*Lo8{E~uS}$5Ky_zRTWA-A8i<J`hFR%5@%)hrI
zJZ0^@X}7*dFN?Ka`CcjXS6@u<iG5DYRVQXnI=|;<=Cnm~_rKEoIB`cs%#3{VQ;mTS
zfcA^UvuoK#FV$MNw9Rd~eZ=*;rc+vaLLsG<ZHrjr!>Ym)=6m)hEuSrPmjmQlT;O8d
zrm)z(`F6{uDy=C0HN*DKj342=e7`-Guzfw$y;twcww2NGLP;y7H}r-0YrWf@RV;Rc
z`J8A;X-Tk-=!AXq(q8w4t(M(*-1Eim)=l?|+E=V!-nY!o?ZlNBo7~upJx7$j9`cgh
zzI4+SUG1L=v&D~}>Hl{7={@d1KG~2`$wQ5y1v#r#@_q;I=il2{Au?4jW`F9Q!+9&#
iFRfPWS|u9QR`pT4{8snt?J|b`e3RZHb+`Hd-vj_#@2W-s

literal 0
HcmV?d00001

diff --git a/arrows_wide.jpg b/arrows_wide.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9438967d7ffda7b2f291069aec782feb1deda3fe
GIT binary patch
literal 25217
zcmeIacf1ot_XnJNFTHmJ0k3q3u-PP=a;dx7Y<kb80cy%-vniXlsenEzqM)GC1uTFf
zT}3H^6$>C(=nA4Bq7*?C5gYPO63RtEpWpNQz3=D!<JtSUo7vej-#K&6nKNf*clP2Z
z7Z35;XbBC$tE#Tz%>sYCi>G+EnL}|do({zmwW7fSo($6%YEJP=N1GA_uL}HK`n`Bq
zadU{{5_J&7rU!ciaUV6<ACEzKZvuh`3n1Rm5&4AIAECHfA4N0lu%6#-e4}SA6ByRh
zi5LWign|k&>SBtr6^(X(G2)j8dX5<0Y-qkNA4|k2&Rd(0McH&+ept^kcpXTWvZ0=}
zWfCqjtS4Hkwbo@Y*DB&Esume67~~g71ho=)Ff0;FrPAAKg#uUr34{<V90W`2M8Z0O
zp!U+yb9gh58k!0O>#R6&sXA~qtmmb6<#M^fInm&FiiTjhTwa0^3I_p-LFod^dGmwV
zbgv4?WgMJJ`%_GUW8!RW3D@h3XSiWKdzLD?a#Y$C@L#DXkx50%wFLYS6{TVn%cX$~
zVdx4tC_6OZuM5UgF)ufqiFs)X3MVLfD0DUVY8@p~j^UzIxq(;eEw>bna@XdTpi8Vq
znUVzRqF$CB)-yjSKn1;-DA#kiImNKtP^dz3ZP7~ZwTNYouBp62RZlXPLCV$E%l{{d
z0lT;&vg?}M^~4McsWdUrq`$LO8e(GfwbBOV>I}f#0$$FGdpT-2FeGG<Kr~1M+hJjy
zSOjcokU$K~X(;p;+Ip%e--`#BVBs$)k}`_Oa|z%o(VvPdQWRj|FU7CJL_NyuVZ>AR
zcsx41(rYx)XeQ02yj(n0tFR#zCdvSp$o_<24`pIL*2_fGm-SaMtbeqslI0}4sWfFT
zB&cCMtyDUmN%^Vzwtdw^FR808#{{_0a2OGm`6_AfQr1<j{#qzRF?5I<E<_|(OI^zQ
zyJDAmmDa3OT`hMh^Y6-)yWwz19)`pyYOzH;S}ToWD6K@Jw8Dv^I3%^8X;Kx+22}E*
zS0&GBa6`s`qGC%VYDD=0w0vMFpf9DAXy0CWbO>rG6$Xa|#UQ@M;8aF&3oN&g5xL4n
z<_SOuil7FYGN16pWC|s?@Hr8IKdMz*7|iW-#A!?hSfWvF1rT+k1`CS1oyMYH3EQnv
zZ~+b*>A*E=qs=PNl&`NjnSFqX5;Tem3b2M$061`50xTu}qr>VJVTC}csH0rISO)Z{
z97T5yWVlP5Vd;R#nky3p%S5C*##OHAGUuc}8uB<x{DjI(kskQcHIuu-A?dzM6sdrC
zRB*;qN|u|C<-m2Me4Q?VDl`Gs5iMT}N}MQEfmk#e0Le;|<>i_qKDC4MmXc7VW}nj$
z@&h6aEfqo|0hJ@{1qlo<U7`km)M~So?&GEVw7~;gZDqjvneIOk0T6My`Jk9jgdu(~
z5PH=XTBkB3G`JEm*=gFQa)^8?hajMUEoDHU-IUv<jcAm{tj~pod@N|U${O_xfyG@q
zlx$hsbV>8Zuo5wb-JF3D=P%#C^jqQ350~vpYb6argFPZNGJ=9WTrr7175K-?iVW+M
z5kO<i<qW(?0%I><S$zg45%pfjBEjv7ahK?>0Qw|=6c587506{}iZL;uC{9d*a$)Z^
zpbA?_d*%>0acOzP<1&V?0D?SF0Q6<jKzr3zfgg`&^`d|%P(TtMry=7OmNj1{`#ah)
z;B^aJE&6YzE_aUl9A?UW7kB4PcDGP(*JKPfOzClYLIG!9pdreN3;TrmY`~?V{UR*n
z7Lv5yh84+3KykU&|5pJoS5o=?lO5^_ZPXKU6g*BO?6F}^uhSv&IuW*PyD@7bpo(UF
z43;Q!2ois9gNnQjG(GB%Ie>CV+EMDa!f3`*0KG?uM7%C*!Y35dUZ<EY1O9{cU(0#<
z3KNz%ymZ5!FpE)wkYzycG?xK^OU9+Z3t*T*kD^9DTec9Q?1?LdFoVM3Mu8=wDB%wx
zLOAMnUG0{@(_cC2E9FwA&RkBg#)PL($+Ey&Qlm1$e%5`t;H7>^q6!){qnHI%Sx^d9
z7%DwQ2Mis>HK;L%N=shdYQc1<QU;tb=0OQD>ds+tR3$+DG$umTIh3|2RH!D6Mo@(V
z)y7dar--1s2%4f5g$jpyKbl2x3^lmXLJqf}MiLd~a2hp{s3ZrRo!O1@b6}vd_%Ym~
zw4>Gtrp_s&s4a~dXr&mni!rN(FrW?%Rt9uh2pV<zF&|C9s4IujIh7vuXs{Tq3ZdQz
zmbR!6)UU(}7PS=(L@=>Moj`*cj8AKDl+Gy#R1-iMw?d!Oh$|dMlnPr;OQKPM!jseH
zQ8um!p*j<qpcQdimqt@=Mb4tvqa3Laqxuw@C4tWd3e3B4jm5yBMH;u{jAj&$<7Ge(
zZ7iT7G0xB?Co0k4DcU4KWqw?cGt;20INxGX0Si!SPz#6QX{8OdIxv+%>9<(<m?o!;
z(Y83IcPsO#-ChR7OiG{vJ0G(M2sKKkFngS^SsZT6=_dj?rwa49i3IJ0F+WMb7FQGt
zx>XqMc4CZQWy}HDL?bFM?U7>foT?1STD(~-4fJU7g|RHD#&UiSR!FPOmVg-(D%Am$
zQej9$okA%HlWWw7B`8*)X^lEZ7Zggj#z}`V3XMh+rI}<IP@xxV1az2HnA2J%8i^|G
zajiWUjVN3ZZ8#T;D132kF~>#~!L&|kiL;8ZK<7XcNkv?#i=oMkB2DT<XsW2lrS)Y%
z4V^|5uu|`}WKab{>(gjfqu?v`d`r%TE8+%AF7L<Hq#;ZfSloabgqC6vw+M_{8b)zP
z+!#QGCfpk_=5ivc42TD5lQJjH;E{;QgCY=~NSioXYQr;Pv%(^a;sT}FMav~P;x?yq
zkQwLaED9?cQz{J>w-w_nwP{Pnio29%(pm;o=D{B+v-+$+b>6tOU{%MJwBM!;X$fV_
zVhdrqurh71iSq_rSuofwc`%TQ4fX_PCX{@V#6lL9P)0~^$X01Gp%akBkUd9OH4Y=@
za1t)yq4G{SL7`4yYcV1car!W?nMi4zaL6Yi@&cE|8i*3&h$|He+6c(+R^>yeN=3TM
z95Nv$t1^;qekek#NYdlY$E+%!-y^cdl`1CgC9MgODk1jft*NvsZ}FM(Tv#QJ`_iGT
zSB+}?x=`MxR;T^(e9=$_RGZ0wIs~iLE;JCcif}cRqlgfKs$+2~ib?qDOeCnx%lK+x
zBpAi`keVN-2^)rKR5?1Pz?B*^8d56=jmAZW%7Ag3+N24l8Erx9(8M)Ns-O>Oa`CW9
zVT@@IOE|}vb6QLmvD>UNtu7rAGNe{(r=wnl)1~$Mqw<2A)dGJWDtJX@Ky8L*Rf>RC
zD~hwpLeQ^61#xpBl-KDraaa*i>FjjETZjd80W^s+aab3lld(d|tjlRqMuto15J5^{
z%Mp4!k}d=Km;$3W`Z-(yV|tgHOBO^SeaM}$D<lei%Ab)kvWOmzXCpR9t%t-pqa8~d
zbe5d3sI(c#c%BZcB!-}-pbKj&k%J+TF2K0KXb`xI!J<iQfRqAV(Hb>s#R4Ht+Kr?D
z4i#MrV^AP87Ci-HQY@6%{Sl)O6|s2GWx_0C5@!r11C5k9#OzUp$t6RwIE$FVZb<-7
z7EBqLRBz`}CP`c>!E>xxMamL*F>JPJ<Zin#WTsHQ21jUf66K3=S;#C@Lh&%flmRUY
zBkEBqVir>X(~(5d;>}~yh$d%YwF-{Zi!8+i9wg1E6*VfY0-MfiK!6=P99B<)5JlXK
zH5O1M1-_iM=unpdDM1jm0sn0iFjku}pjDI6u+1CQ%9L@DjfHjjQa6E(`dB2(*cAzb
zKT?p`En1^3BDC8Bd6SkzGWJx!j4I`od5Rq|nnezjCDnu_11nXe1GC0pb&L!XwjiNb
zlX;`v1)Gv2<RL9lo7rLH96G`QJN#NF4txBLWYj5#14;*?b(J|36Li+8hTR1?;&nQs
zo(vIJgCB1qnu47IhcD{LMqRkWABq;8E?Xc#5hATCl%RZsMDEJTgKh`E=tdp1L#U|u
zL$^5}vI|v=JLq6+PMzPK#ljY$$>rvA5et}{c}&r$+3B=-s6fo@^x8a`5Nmc)b`L)q
zHwzi3*OX6KgfXu-NK^oAs$|fc4W#VC4C{sDX@{_o^;trkTO~w%A#KJhlq!9BBul9v
zt6#z8LM%@E?TLKUrOx>yTp__4Fh2|zGinQ2ZgN1435sem9&kHgG2knZ;Dmg2z!{Jl
z#Y$HwLmBdjj*V$4Dj>14$tabxO5LKYBB&r_w5u2jl7u|&5@W%bmR|<Uv9fSbY=sav
zrlxfaric?M+AmZXG&&odwc!r2NfH8s3$3w-LnK8c#corG)u~{yUm8OA8cY*PFb0X%
zAdY#;fJ{)L^Jr2!rl`<I<5^i)Wik}vf_&J+7$Gqd3#TP!W1LS#P_M<8z?~60-<s5F
ztPwV86C*}rMCv6qTAMCvwmHgxF2tn;mgi&<zb-0ry2RR$A*N?MT5Ze{3l_XyZORc7
z=zJ+8=VvuBzZ`)hY(Pg@w2};)7X}#wlEhUpTBsu`h8_2L8EsPMjOV=JK++sb5Z*{$
zLW&dKSWK1l>JwQZ>q`ddB%zDvbrDh08%wHnNpmu%O9i~yM2aY+1v*%n@-dk*hkCCh
zl*)Uvj1)!FYG+Q8BItC$Td+y>igb}Frc#zDr!BxrsZ+<%Dk0_b<v1ZD66-^rj8TQy
zQmi5q&PX_4x&oMyP%@3L;LBQ_@~~8_%f@YdNGg|PAs6a5;OU$*h>2y|V2;b;PJ<aN
z8mN`Uv_qfwhgH_J56=tK>a2{G7xcwS*E6MKNFf~7ak8|qAoUrH{sLUIyNsM)EG(vc
zW}}QR7T~NU!zpC~pVVfS>oI}AZqIX8Eo{g-?3~LAM_o>#G2kyZ8Rp||k1-q<I*T4z
zAR!cHvp(9GBSe~1Kxq^@MWG@U=VS?y1fq=@g<MS9Lq(I;F6OdfH{YB>)aD3eaw-v8
z6e|NpGazo3uyI3%kx58jLSRbxBpFxAZ^}xg8Xu>mgnlU#&Tu9<DwP#;PBR{rx%7Dm
z(ivoVwwQn{0=YpClV(>Xa*)SFLOw+4_!uirnj>PqKPYiPDV8q^$}p(lhHR`nlST9p
zSA={)2!i<H(hM`=6KbQV%3g`PgOCr^SSm3#Mq5C<ODZv%BrrOn!YppBB1eOejn?C6
zT&Kvzb)|qAM{}ir3yzA-N>paTF@v4JEOuo+=Tqg=esx|JEH}B7DF5e_rBKd@qXIXm
z!L16BR>B&u+`CNk55Ly})+2^wV4254gUPqTjpmeA6xCQ!i@}NpAthRwhO5z1!z}e@
z{uZJPh~_jZFnLic&<JQAASeane<rq|3eZ|062C@aiR%HC0#pEkc8db!0f_(%;KnL|
zw808kAv9J^-r`WIltLxrcPmLlLKVXD&b);&V<e{z8S)-Rrmz<1tTJjTTJ4dfuHY7#
z6-Ls+MkJ=9F%+oyLyy{y3o;swHg3==LbNBO1)@vZ1P;#V&|#^97&0qilU?9(l@<t`
za6aIQ%AFoz2y}KX4VKliZjluZumLeu8B-F{pc>`}MJ8p)?Nu|Hn2r$_fT8mh5ya+*
zCMb8zlk~A#A;luPa6;#dB_vEDnare9UP(4yk}qh!2FEO<vfN}~CQ1dc3SUXUAtjC}
zN+E`lAg#ubhW6xXkFpT=sdI!uZ57*$3c{gf{8?pDi(7Fz<h79nC*uZ=1QVGn>o!J8
ztEBS*GLoaqP1c4zgo+770t6|NB~8U#fldhc=3+=|i*N=ziI}a?0GO*Ld@cf}T(A{P
z%>!|-Gm-LDrWJ6;C%i0!8#W)0YQYeTFrHW<<7YjRU?gdZB)ncWsf#6(k^+}X^0Sb)
zqGdD(%mh^^OLm6KP&p}1N(q<y1}CS9?+X@jS(H!4gCq<&;(BAjWOL@tCUM2v1QDgq
zpHP^osKG%gseqa9&-g+R$z`>|D5DL)s<<R1v-1g(w#ckvFbR>vdArRC<#3akjroc$
zqzD;Bat-A+mN~R%j20@F_QjB}%_^~4qfA6C&1flu*Jlr=!4saKgVL#Q#BjMKPBx&l
zvH`j>(CZXMLq3jT8Jo@Tpj`YwP=>OaqAX+9yOI{CTo~}dm@%6avJsn=h7dnuPT(SO
zlriIk(Ttf*EM&5~5QcJ_D~3%A3f7=rP78CkqSlymKshvtc*734I2Y4-DZ7ivoBejP
z*>5u%%^*gLq{9U33B!7WH^LGQGvN}H%7_D!KxGXU_7(L>t6HANd`<}qg$>e-Hz;GQ
z44q9t1fF&TnA8nVWW;OJ2>eV)EW#Xu$~?t^VM2|@q|35qnK%oYt8$uxusk4R@(PnL
zAZ0Bfiw5#@-iTBe<m-wYq_<18q>%5CG9s2sAbLt1_Npj?xfHY;m?#)!j4DUes8HZA
zmbWF!b_0i?B7-MH3ChNKU^Hb-rO2QalR4;+H6)ZcV&))$Q8F=jz;dMpWi4O5yppnH
z5|UIR78Qm?v4qHPjC!q*FQaFy?s#4#0!vCvDxWQa*2oYCWjAV+Q4aBkg055<P~cOz
z9W+On`I#J&wL@U_BpyW4q=4@+o7HU488lH@n?#h92qi8*t&sV=ggdH8rQ=$4K5OxT
zjuW<fIRlw6Rt(!|l=|`(n~2D2al)UFfG|N5)}<LVO4*b&79z={!9f}vK`#V>RZuYV
zgUzydKo`-(<x;(0gyRsKk$E91?ar4uw78Pa>l&^j9f|n8My)1iB?xB#EX|tzEaB8i
zV7||ccqKX&WK+jXc`M9^2|}l5$jTsPGU2cq!U-1}%a*#6lZ~hfxHdsVDG8@Dr?Yt{
zn3%}WbV64M$|5pSAme1kAP^rE(<a@N-sVoZtQwUNw^l+CiA;=1psd!LkZZ!g=d)2$
z8d$wSm36T&>yYvhSSN7{98$)LI0>pCW4yRfrE~fydk~D_5Nar74M;fZvWF|{hYnhf
z%Yh3>qhdCbcHlbLDu{EyCVW9lHll`5Z9XYuD5qT=kbrm;WpIa5q~ruNkqLnw!ibUu
zNTzbilkPZImLZ#E@q!+4%R~}gKn-V#PA`l?f&k1{II$#&=Q8HBGU%s5m`bQq=cRlv
zglK>TXdED@vT_z#)@HPcV3RSJu0*h=f=dzROMN1KLF5a%^*RyE@f|8Q7>8t>Sx=<w
zOimyaM*MhMmUc&ddC2P*r=5<gUb9F*ISdbbqfnVcSSK?>z`a2ZJz>-tM3BiH%P7Nw
zpgxeKr6C|X2nHjUAPVQ9ViZC6ZjzI01xbA>BXKEXB45G)?nXE%xLXD!_&!HYq48w|
zX^9-iz>-4@HfEi27a~q1!m&h<4e))cIL(M<MHl0cK#rIbmQgGmaA(zo*PTc@Id3uP
z<0q1pCPRK#5l$5|9#_mrMZFnk-l6sy5sgeN5#>!0uQuo^m<3UqL)77fIGO`%jR;G)
z;wm~=b|wf2x*c9Am-5*(We)AxLRJfVl2pM$rQx6_=}!3s(t@bu<3fCw&!5mEAvh+6
zu}}uidNmxUcIEU&Nh)oQ=JQq`oGNg9XE0(eaFw3ovv35Lat5fVz+vF>*>q4ONKkf(
z$*EJ>FM0li%aQWA0-W9}B5ih|pb*u@Jp>+?q;-UogfT4d2w*r-iEVH-m^|5ZT1&*M
zPZJKS1uS2>li9pOsLPa2S<rz{of3v}3MdOnqaw%yGcnkyhFF|}Q8Rc7Eb_~lD5Hcb
zp@@mE6UGQo$RG|<4hJNP!xDZ(2#K5#19&|kqcx@-oXKa0SS2J&iE*dXMM_h`SVohF
z3SK4*dZtAJ#vz$lUNM^}<`(EEmmn6*z<HzJ4)mDLYDg(xr33R)8>Dqbq&6F#ph7|-
zQjn`@F76hlY`_;K6?P_Msy{rGl~T{)Uh%O1X<ixde~nTPpbU6rK&W@)Kg}xx)|bJ5
z+SJRulmY*C+`Uwr8#vgQgYV^o1v<#@H3XcT$CFY!T(NK@qfsbiT1KHXN-7iZL=i)z
z0&iR{b9nR!B`bO{BrJf9Hi3!HWTiU0+m_>;p@fT(#KIa8lS0)5S-fJVWm~;`Dsy<r
zelO>hFJ-{6kZ@4u7Zt#;fO=BEv5LcBC^xfgA&X>W1<Ju$U1R`-70z5HWHGsmGMx66
z0c9SIFd0rs4LB2(8=xo$+>kNuOi|H<h*cN^8sKgfMr%0fGg&}yaXC|#2w(1y=NSfg
zc%_WUD<U!_Q?wM4l?XN*4Qo73Qc5RtLJ*M<d|fsI1}lk*FOz~_usl$drovD%#l+a4
z91l_fh@UbhSdj}14}=2@LYH!LcqN=GdOaY5gy4F&Zdl--Xv<BmU&B8Y_$S&j;EmY-
zPr%E7H*9JQ55__u;Ld;n9ZSeeNv9(P7D<Y>tkUO$b!m?!Zq~<&a+6BuV)-RsSDDe5
zG?ztkR3NAk<Xkps7M2_Jj;J7mq_jRx9JP`vV>SSS`iw4{49mD|Q7Vg4fvhcE)bg1?
zASezBJOP7Q6s9UODP=a7(#yCQus1dfOd$+<g+PZRB1n=o#ezXU@O5B7^aY?wE^0~@
zfsRG$(lB`?w?bQP^8cj3Khc%}|5I+R2VMre5&PFG@K3a5!0XG#ZRC5^h|w$MyZkD)
z=uJ>UPMC>vx`>6#lF0-E`dh}EfI&1s1<O%M+=X14nngm0P$tVn^F9lRXRNj$1A8Ty
z(?F|(0tX+~m6XC)=iOkumc$`FD~mu{a|A@9DVvC&_2w|X40cLTuQOW-I0aFDQ3*8b
z5Xo$Qr!ft*{m(8E5kN>$1(p}H(I6By2EeL^(w(l%q(WjGG0Q?AN;9eAih@9qi)2KY
zI%N-fLJTG?*f^OEN|}9xUP3{7UnyQ>WGTL@NCf$wnBB!WonZwQ1xBUKmSyNQa!8QX
za!^DZ=cHsZ$%K@s!$A<F4@Cng#)Psrp6p(o1v^zxtg_6)2kX?JnS7TjqIPM)d|pDz
z(kyHvGHA>YxA<NDppsDI@>mGCz?3$i64-4<C79#eIZaxe0#A>ijFiGqMv<*FS)eA_
zj6LIbSTmK_;~L+6J%4`VymFKO*C_P>%7Fh1hU!68P|ASUAHQzgtiPaK0mxR|mmZ4y
zePouOaD^a=O$4F9@#(z4zGX-xRUmB?2_D^fJpoHH7{|B}KTCM+3L!8<T$VN3B`^}m
zh&efAi!l*lnL}BSQBr)JfF)QE(jnepG-b57h(t1)6pA!iTp?kynMl#dK}m7St_heR
zq+oz_VC^);XU$1LIGA=3S-({v2~^@kxgjBfz@x^PL*nAQ)e)b=78gj1SVAt41~g)6
zQUhjWSOE!!n2<SVv62D94SQ0V!^+*DF$h^etQrKdHSW(qQd61~hP0qdvUtj2Nr8ng
zmiA<Y7C0%PAp>Itp}5y20{4pyh6tQl*yZFZ%Pdi$Q0{_3_?ZK*vJ?QB-38_<IuVi-
z!a=Jom-PnBmJDSKNP^Og4LE7HC}~EN!5ESdYgrlR(u3(hT55Hf#g;OMW(|;_&&?*;
zXtszaw25RUDxz36mx*iLz|%y;NikTBG9#Q!AkkU*jM0?^PoRtxSn0Oq^>ST22V!;!
z9#=!AG9XxUNqI?!Q=D}hogo*PQb;hHHz<*r<C$PW3nGI=HVR@Jq`((AePK0-q-<2t
z7Y7d*10I1n<3`Ljk5v})5k%!(5htX}A_>Z3HfpQ^0#1t9j5=W}MtlMvlaMM@E*jy}
zN_X7js6>IuFy-RFi$@U{q3l|pGp$hBD9S-I8m97GFX4$9!Fvl1hbG}h+>vP5ZB}dI
z9+ldXw3149@&-LA5p_htOAz4stR;}M_#N(I%7v7h><&heR8S*f5i0FTsDxgR+rdQ5
zV0{4&i-8uyOsZHcVqncB$ae*xLeS=s7OjX~9#Td?<YkVfb6yK@u$m%K(Q;NFEKNbR
z5G%FQvJ4|~q*8tz<Ow^nNdW>0l?k~SiX<4YS|_J~XZ4u%a1q!=*6w5NiDX2hXN>V2
zY?2tPm8qGVlCXqc50+Ov{Gbbo@Eu__fhFzUuv-iKBxQ;_SbJEkLo`tfg*uXyP$%Hi
zkj4zw(1MX*Q7Le+DPvI2Ph@2kv*AY!Do*8tv>Ba_fDj30m9ju%7PHC4LIn=72<&pb
z1g7+AUqQj82$e;~<zQXN?ZEi@q>)ccBW|!D&mzgnOp0|NF&QMXh9cUu5oNsKJ$cfu
zOnTKG1em-PipNbkNhGCXg*e}ti9-p!EDI**(gI>OCb*<W0puK{Lk<@tt}Ldh38rBG
zhf#wx2PHwbH{{KFU2=|~oM}3t^%di3@WcnwS#u7s76<IQ@<<&r>0J?@Go<x<QDD=-
zh%%&+6rCxfG;g46FjP>XYFS3-3+f>dvD=fP2v{WLl(5YSCXjJjg$UDyLek+&kc!F^
z@HpsR+5i~i^R@_hBpI+L2({55OIq|f+A7pz%A`GSQ9=q`jKi2@lu1b`2+0><wM5D%
zRg6FEw88wevnYk3Oj#+?5FZ$!oKAriNSll_U2(wIv*?t^<BrE!@aC6W9EbHbo57L<
z18+jB6{2XqvgRNYK`a3kBLc7z9d|`+-jpvda%S;_nDCM66mYs897<C~cOoGZ+KWYd
z(QGzr1ZhEv08f;)o|q?0ri*GnSU=U3Ib=&+GzgYiOd4UF;`lCp!f9858Fs>yQ`3>O
z!3w4iXn_L_@TdQw0cAkWW=&|q)`CtQ_7g5^h*YZGG%Zfb(xv&7SOi7$NCYVgxwJ4O
zp<%m39<sWbpe!RYOEWA#kvC_Vn22Us7sQqURbg2^s23$&DmIq_!$g8)oOw*{*O`%&
zCkv)nR+)?x$`Uae{81{(7*E6tjtHn7yw?iWFGRo_+Vc*%1}L`-Xi}Tvf=tlBXAMCv
z;dLq95RB`>@~n&lZz;(Q*DSJLp2uEiR8fKD_A(&ir)*iHKx=o&b)gJGnF)Ow7f8V)
zWpPoPr)Zf@03M4|IKQ;0Zv}6Xz<Rb2Owt);iu4AP;1vZ_rerOCr<tzI^&CN#i3zk&
zF$^4-OhDP8qBCRE%W__yIU$Qe34x3Zf_VbYskLgo(`0d2C3=-e=rrk~a{E>OA$erj
z)z`r*?|T3FZSeoewusV3fWN*&|9`nT!eIFOT@JArbiExCY0gSr-ScwYjTPzAtNYO9
zH}?T~uj>OTvtYB!-)*$G7D$-|J6*2spegM{0a7v;>NISc^Rj+QgAWH8gBd1Jr&K7w
z6K5f;5aUv%7=|&S98*X{A~}pp6hcL1pTM<6uWIWOEFSk~N}Fpg!3O>Zu=R@m1=a*S
z6vKL>^=L8~fQ{fnflMwG%S9N0V=yd-#o)1lT!4#UL?Bbh%GL!~zE1hS)M2IA08`pl
zq=<Uc>EVe~JPvkM1cH|_{&fE@<#8&-WT^m=ipR>^Z2oye#?Z1|)G=38QErr%0?m+7
z!9gN{KrqM)i^YTF6eSuI2>3(@BBfxtgc=I{rR-IPN?NYm4*XReGL!&oQ*wnsDHKR#
zFf5ZPBq%IYD8-mqf{O9W$n}a|4>|rHYox;2b;@6_8W8H<Kq(cNKi$6yaan8iv{G+V
z%_VD9Gik6lr%>-kNpz*%G&U-EZSLh8U=P>-Xtzz813k9R>n|DB@U#~gNJ+r;@~*NB
zs;C@5c_o5+cJc3b(Z3YEF@7#Yv46McssdG}x-m#P9^`V~6ot~D*WYO0Qw8Z-(i=ni
zLtuj(6&Ma(mel2pKNtDCvEf<;UbcomDg6IzXkf1H^t!%xu18Sbx><i~*3f!4>uj7i
z#bpxaR6NLl-G!I72L27jC87MEH3fLQ>o)=10IZ9y-NOem>j1BTjGszf(mr^cUK?e6
zDQ~J!D<3QeoA<6E`dfZ0tpHm^!7f$k$~M)%^u)^@<^S~e&(`&S3Vs>vzm;5;#J~0X
zUvd3gV!BN9UvXWQ#J~0XUvd3gV!BN9UvXWQ#J~0XUvd3gV!BN9cj9Vx^+Ok6J}|6j
z4t&Pq;+HL4;JN!9W~+(N=#{*xDju(eTF<cY>NY%HjO9{R6;|tVdukhh$gAcx<u&5L
zJf7E|P8e*2z4W~ejZ#sY21KRw)#InV-~$}s*fmIPuC2YA|8J$*`V%P*e8a;Gs6_$r
zr3#Rq0MgMMmnh|*0qG9DNGV<2pp;Gl20{9!Qkt%$Z!6PO(j!ahK#UClni9{6KrB#7
zzXQ?_W-}B>*XTidVwRzDAiW!;dq*=d2Bd!m=?*c<o96Kvv?%3sls^R00+4QzvRf4(
zJ(S05(t@s+_SH*sR37lj<0;~a0+{WExZ3{y0kz;W9kN=me<@0F+#oaf`iM6b0N)0Q
zCA@5b$E&otREpQGB->gbWC@tCN(KoB1My!u{`Te-s=rT^bX%b~Yboojs?+wXc~{F`
z&5JMO@nq+KY#zOu=X;#TTX#Q?chml>dA*<J@mf#d@z#D+uRWbg_EPV&B@E>sToUM&
z;~xpGP~JFEufD;h`mV^X7O#9Qq_!kfe>@64@KBouAAg~02mRTI|Ko!7*s90JZD28q
z0`qK8l>=BA!_q+AmuBU)40|ID|A)ou@lZ-u>>4E6Kl2B#!ySWpZ9eM2t2wou*PwlK
zUd`wy!JVordeb*`@Jc<;dCNByy9a4-zkI&tRQ(9JtWGm@Z7Ew}wb%MHscb0?`b6o}
zkk^dYn%9BXh1Y}Ei`Snwh$rMpc|&+3c}kv^XX4p;Zr<%YKQF|K@lw1TZyfI)-bCI*
zyhnL6d2@MB@t)x=<*nka=B?ws%6p6V9&a1(W8NO#0p20rQQk@38QxF4->Ryr8dtTf
zYG2i*>XxcLRs1Spm8@!bm9k1-Wvy~o-C0Ffu~nI>aaH$KJybQVYHrnns-;ygR=ra7
zR@K(39aW!J9j-cF^+VOK)iu>EsykG7ukKq7Rm-Z;YF)LxdQ3H4ovbcaPpF<+J*RqM
z^@{4X)o)dAtKL(6sQP5}xf)(g^O{aIx7G}*k<}<_Of{aGU`?v#uA0d;vuYOBtg6{i
zv$<wh&7qoaYkq0as6qP%wG9R}7}7x9fNbDzkZf>QgNGZ;ZLp-l+6M15*wx^0gEI~O
zXxOr0w}$+NbqzHQT@C4mxrP%P&T6>0;hKi;Hr&(jXv3cxHEPtUQNKpgM(RfHM&U+d
z8$I0UiAK*idb81vMn@W*YuvbTm&W|Y!y20!dmD3&CpLb(@ruT8G~U_xXyc!oG;eZC
z6JZmgiKhwM<lZK;nk;MbdXt?^zG?DX(>6`}G_7lDY8q%-X!=OgXPRzk`cc!PO@C|F
zwpqVs!<yNfg`3^eY)-RP&E9SHS+ldvn>X*-T;AN$oN4~A=5w0A(0oht!_9wb(XPe7
z7K#?5TVz^HZSh=-H(Pw#;%v*7E&H|{+0xUJYdN*$(w1+x{JiDQt=hL5+)C3b(CY41
z^IENKwX@af)=gXYX^pnNy>+qmtky5J{;2i0ZJM^}+eXpG+vct|^V+=9W?!3gZQHkn
z+nU<O+CJQNS=+5`kF{&uu5UYKyFj}M?H09ryWNrYHSK$~SG4!Hzpwo>?cZ&Gv_qo~
z{W@qmggZRcVP%I8JDlm*uA{i4qhr3~ypFGRJk+VCQ{PUSPSH+}bXwhMZ>L{7ckhgL
zraDjRysGog&gXBs`KFOKQ8!J#>G_-X-1KXg+Ahj2kuHyQS=Z%Y*9Kh&cC~cPb$zny
zyIsG%xx>vvZuZ|i`R3I(@9$RKZD2QRx3S$8b=%(UT=yQ`RoxTaAMgHl_iuZ2>M^oM
zxW|khulG29OZ!`f-@@E7<CZsWIZ@lO7Ojoe&Z&K;_ROu_ZdKo!xpl#<AKv;~&wf40
zp7-{AzURSS&3egt1$)itwW-&c-naBN^d8rHdGGywn)Z?P3H6!XXLFzPef#xw_I;r5
zy1vKy-PBLhuh?&Szt8%&>_4J^vj4*Vy9YEHAR7=F@Wg<R2UZV628IUC8~72wnlIrq
z{3rN3Zfkg({5JNsg}3b+)M5}iC_8BRpu>YZ4b~67cksHwr=i}E2YL+J4E-TM1W~~=
zf&*|nSPS0^Z-9Rg4ix%@^Mt!aEk%UrZqa(t_hP=75<e;a1Zj^LkOz=W$OVa1l9H^F
z9GCW%-YK0g-6!iHGs_;9ZIL&WWAeM@ugZU^lhko_tLwfSGH6I-$g&~FhV~mu4Si<l
zkzqZDd50|+c4&C*@H>Y;J^av!TSs_DEFAIW$X+7@BNvbS1|5Kg(G}=vOn{}ZHQ0GY
zonpM=O}qwI;gj(XmF<+Ia<1}oq9;KU%ZW28L^W3Rrn;e8r+!SmPt#rF*DTeX(Mq-B
zweRR!=xn;Vx-azu^(p-ZL$yI`m~QyY*vH5k*O_=GjcK~+pt+wpY2IjQWHDRjS-!Ty
z)^XO&w)VC=Y|Csv*)jVg_5);pGDB{1w04YkEOnfB;?8N#LoUcQ-nGr$)y=rqdm4Kj
zp2ePXqm-j&jygIT89j0I-ZA~g<i~8iz3c7K+h4n*^&Q?jUb?g4oz6R#-TAxM<Xz}J
z=hOJ+`_A|k{@MPMfsuh3f#cLLY8v%Ta9D6!@EAReo=zVRjR?&Oeak4Axy<)rb@-|9
z`G_g<T;yWZ8GSz1Bz9-)6}BB4Vc&`Oi09)w68yx(#Gzzea%S>ON|#!ku1b$duje{|
zU}}4&f9C$o;q372yzI|8NA9J3+dP}!UKm)IRQRSy6rUa2V61=arg689yJy_t@#y#k
zcU9fxyKB?kz3!fH_tAf;{<Y+uX7@zy`QTo`z0>af@jmB$8zyw0@UIC+?pNKvd}6DK
zsfqg@81leV4>ov^dGNzYqDhZW`eSln^7{`79-8&g?+^PQet!x)WzLj~Q-f1Kd_?ld
zlaDrflznvXV<R40GOg{j;<O{v4b#`osGTu+#*Z`an7MV9c-B+1o6XM5J~T%^XZ_>7
zAAj`m-{*$r?wN<pdwzbm`IF|Ke<JY2&L>Acx$3FzPd)V1uTL{if3iTeVC}-b3ui5A
zv?#mi>t`I#Y<^bu?9#<u7eBQ4!gK6%hn85Eyt7ogbm_8g%N|)?y*#u0_=?dhK3<8h
zT)%42ss+z?dj6pocrRpLIPv10FYaBfTfONe*-NWl?(_2eHSO0-UR$-cu=dQl(7HqG
z-RpO}qJ3r4h9Mi)Y#hAtxmRy}b>3?oUwh>BX0K0p{o)(NH_pA8c=Ox0=(mn+x?|IU
zw_R`Ve#i38hwtj%-SQss-rJke&2MZOw&m5Wbz5I~U-tgGZIW$kw<Fuve1Lqg=0nMc
zYd?~HwEkoH#~XGG-SPU)5j)@7h3|TAw`TXYJ;psd_L6%)**ALMp-%#z9Q!o->G%7y
z`+qy|ug@BOHu+$igR?&G_W8mu27K}SA<3cF4&#Tne`)*jz!CqElV7F3`u*sHuUmdS
z>ziA?S$a%(?A7DM@sCf8I&t)5;^c4NPCV82)cn)^Prv-#$nUnFah*B(ed_y*KRkT4
z%h~6C6#uyCocY}6KgE9f?fm4QyZpT57uheHe|7%)&2PEi8~;A%!hj1K{?PsL*~Qq!
zix*GxT9rTG#S6R%Vu7lQ2Y3^BEt)iG(zHp7rcGP4ZQiVT+m5YUv}oP&ruOYSwr_t^
z=XM>y(XMmno4fVsc5~<6z5Dg+-5U{!M97dUM^($_&0Dr=*``&iHXT~EZq=bfo3?G+
zbm-opL&pvsy7%eY85~{v)b{GpqerjW{{8#(>EB-{1RD~rlpk{OO<vn3Rgdyc)>QT6
zRky9GX<K#i9o{1#ys56PsUPDsYTTe<RdtgZ5MY!-nR-W68(vinpl;HtQR7C9%Qvd3
zYZ^3c+vwJIjbSu-&w^>~J6ODry||;N(7LmiZ3fv1^QG=xxLTx`{!&Lt=eIWb)4T99
z_bqx`Y)9Ju1s^c9pVtELQr)(uEpG&G&B%InSCUoFz83ubjettBcIomVcOOum?6diW
zmGSB4U#NAiTse9fvhbxh_sh<=V+7O4Q;Wt;{NWyY@4g+Q4!cj%Q|I>nH6`ylzv<Ky
zkIvh>H#Xzf9qz-K=)7H;hm-x<{e1c>ezTMJXEtS)ta?)Wxbygyqp34}hdk)Ne-i&C
z;?C{s!gY1~pX=QbX{*|^@LclF^J~`gE#qQeJhSZfhV36c{_A_}+Lu~n0`3vP&6*24
ze`@;lUpNPLaCUB<`I6hU<&%Xz-!DCu>|L1M!`-6L%@wZc6j-P&K2X<X;k$P}F*h^%
z^>p&BEsf@$TXS>6H#`G7WZvg?ocP@}{SV*5Bj+vZjweI%Y2U5v`qA#=2Nx|^{O+9R
z_TE)@^5C2M?)c&8&bOR<mLGYm`y<LB+~%Hre;adQ-r^Pgmn~brYGx_0Z9iy+?Z;OZ
zP;X!4^_Z)HPc2(L;=9>hT+!^juJfL-KKJOBh05Q;=b^8#kFMJSUgP(ct(rBp<CCj`
zwoP9Q)O33|oZQG8p5s$3cRu!G=FHmsEgSa5q4|3bDvj@VY!(0XLgxN4Zq-U>baTh)
zt$PgKy=2}OZh6!3XTy6Z4-2idY<NAeW9MLa(?gq<t~lNM2!Hw4;n(lDefb^c<*j6E
z=Z<LC_rf~|4`&B<Tmo3HiheusDeGpZdFj2|tSb%^@KgJb_M31=@Z}4w%!{(We>h8d
z<gpKo#De}!t`U9W8)V6<)FRKkWgFVGLYFVw{!4aB=M|%u&RM!)_hYMa$g)*miPy}t
zyg%cUXRIGMmcG^^Jb(3%ck1qO?~6_PbyR#^&us4}^>^2_9RYWLem@ai`R-Etj1_k*
zFup?{8AJb3tK4NF4Z5x$ko?+7(!gU|-h1xmnQf+wk<6UA1ew?2^s1S|?;n!X-@QG%
zeC*_B5u1~A_S{B_XSN%B@@VR{gZl<9{l0no<*W8SPu6W;`&Renos*}}t<H^p@S{zN
zlc#gdXB<Sv>t_n%bt?V~r;yHA@}@15mpwRX@~SV}-Atg{&yR7cv@d_W@!X!_E7y;o
z`Tf30PpkCXs3*V4KK1lE+wjGP3;6e=rv5hnb$)*LOW$tn|FV@&ZP1?h#n$lHuA9Cz
zc2Mrk<sO`C-LZ4W$o)mN+thsAoFidm={q0WYoO?pH|<%|`FN3L&#6hv{MOSIvrix1
z#GdW(QFL7qkM-8C+q3DQi<-1#@4$XL9-H!Zhf}A|Pck<-`|g+9lB@UDO+DQ1-JVQj
zo$IGxwr796sUH|obg>gtcITQ*bM4cOS^v|J5B=+AED;fk4X$NvcP!jv*8UhJ2N?!A
z2e*tac~bk+`Oc3H{Pg|Z7e);_aI5{TQMz`o41a0P)ZD;%wr60;O^>%9I%ogG@fur$
zAz+qWI-KzX9h-S~9H4TvhgkROzV3Ig;J#>n_o?X@gfCP%bZfVqGqwBp!PD>W_C4|8
z{VfWTw^O5weD1byBf~!YE<E#0b$syNv%S7oA9`kzbmmRx?s&K7bGP0%eCOE->lcL&
zG=pB6Jp8x0M|-vKf4*Tw_?59u<}7`@<>d1tjtx3zJ@U1p*TxTO1}>5XC(kdAdRE)>
z^rGIbU9a?5y5h-I%bwl8N2C#L6v{`xG0FbfTaSq{s$-|7EgrsO)Ggn2+eI5zuh=6!
zZhdmz?_1#uyKZkF>ojceW_om|)nD(;j_RQO?wxO>ud<<^OtTKZ74E%c*_a>GJEoi_
zcHeFuaCYG2eW~ZOD_h>ShyO?WZokViHngyVYVo4t-dzXoR7sN`#kF7kw)5nGbrH|r
zMV<%uF8;Fbgpnl7emap^@zyR?_x7Fd$}L`dOK@1`^dW6*W6w2j-*=i!?HTgvJCoYW
zI*y~ETSruNJrErIq9^gkEy@n9@wcCO+hf{4qH}6qZCT>o<SSGB(GM=!xMYL*t8xxF
zRXrQjzrSL0^D%WtzJa^{^5T%imiLdXzsPI$?1ky$8huu^k{z{e#h9Sy*#6V++unKT
z7yW~$99v#Ew&c(wEe9;Wx&5mX%{3F(tOz3yP6@L;`!-2k#{K{MFR}Ik(x#fZY#>98
zYrnboz^%Q=$n?F^@lD?I>|ecj<zR1cz=C#b`~I-CYwo~j)!Q3yn!0w{h9_o<)53gj
z{ixpIsmjy0ByQZuo*NB+n3nyP_~ZL$`;QrWHQo8<n~^_G1(qB?wWihm&X)P}{=0gN
z?TpR3%d<bW?3<+q4qvx-aFeOzxz@Mu+PCFUj(hIS9y63m+&<dEEnKjbptheK`vTuL
z<wAIpl%Yt+!n-G)+MC;`Tsm8{a=HNNV(N#wADE&UxNt<;xcKa@`ySy<`I*;RdT#Dh
z;XhJ8tK3U}dj1<|!iT^3h2#3JJ9zZR#@o6~Ka9No9rSd^`}T&%FaKrR;qJlb_P;Uz
z_4Y%KKD_nJ+%IyoKfdL*plQP1-(G$FUpEb_+-!S^R`j}S`%_I<o-d5A`r*C6&xZKM
zZO5wq3se^1-^%ED?(8t+j$ZRW-?VI5_ZBCXcRx`Yq`$D4nkQ7c+l26>$%p2=@&PUw
z$>Vif^P&H-56H;e^=q~+u-L}8YQ3gUcQU3Ju(ap6vs>0nWIIM5JzXW-Hp=|UsB;5W
z9b5C`(gl4t?^`2tDt`X)r>!5IYds)*GPC>b1;>A%?LQj+G4GxF`UusF&m|r{x^hMV
zao>O8<%N^Jc=fT@zCN$nye->ujOp-?zrNPmx#!q{2b3$HU;g2)O^vpIPgVc*sOq>s
z{=<0Zfd|e{8GxO<njX2n+roj%diEW-@xgfNwC=U98}~ZC@E8MszwGF{T@LK|qIjm$
z;!#`H>Z&C{(Mo6EQ}!j3m%MR!UGj-PK55fp%}e|>yBrITPI-L_w`lZ}uN>d+`(^0h
z)19WwhW1Cf3k^3kdiRmp1CKv;krzAp?AZ6-DX-rBl{X$wb@qeRYfsMZyI{d1&n}sz
zUNdTP-w9($lKiCa1o0~oYVV1bo_X8mjj_)eh|T%2<?Ri3{xPikLXLOhJ$MQMFWHL-
zA{GITe~u5i9{B%vd2OFRF<k%iBNw_{<n0_H3KitXmd)<8xCfV7cb52<`mF`Q;`=7O
z^_zBmznQY)i|xCGUFHej-q+|L)Bf)F9-s32`~QmFGpXa}dxVEu#)dxhNnl0xyPAg=
zzp@HXlN~!JZQ80WH)u&~r260=k5O-2nEUy-JB~eZ$LemcujewE*taL^q`8G7*I29f
z+_m}Cz}NeXSJ+<o`o*zZCJy;Y+U}33C%(S0r0v1uo`!+kWM)D3WfRAKyN2wnsv0rd
zzpsPImc0Muxlg`6Kl8l~2l{T`ab(9<oiHBW^3@%iZyq_-_Vwtu4G*BHWxl>giuCfr
zvh_FD#TEsB9oPPOZl!6_ZOM(Zoug)axri=oYC*<M_8JRsz4@26KmX8m!~8eKKeKto
zffGLbSL38(pJ!L`w8yg>$34<u*)xluXg_|#_pjv7&7VK<#3Rv8n-9%=x&50PPqv>o
z@UFoRFWb|xqs7+Huv=FB{lfh{tSjDX_eyQ?LYMu=&iwZN-1&=tox6JK>fzI-<xliF
z`{mf%KR-#DWZe(X4fWR4KDI-BaPy1gDx<hTQQcw8#u34dcQtwc8{<2_c5E{3kA6QN
ze7BF}gBJ!*JQy0j`{h*^dFy&!<c;g?zH^7?rM2!2hxQlus`zi%Ml8E)*o1v=bw5uU
zj+Ir>YMQQtVc*PCQ`Rqx$Sv(V4{kE&dCK1E(4X`KpIG^0$4SebhK^XakN@_kCUw}k
zH+qrRf6tez)~)R~_{f6vqPI`KxaD&U=RJQiJ0d#0<APOl8-A|wd}_Pz-DA6_?)!GR
zaF_1{IdRR#tv8>$<+Iu3h_*@I(C-qf$DDfT{_)**Z=KrpBi|zDJ%4DwA26Y4`9Rfk
z&6dfN=L{VD<$?vTd=-5@{q8RF>H*GkFBG5iZTzEe%iG1~HTb*;5q{&-J?$p0Ch(`0
z7Ef;8epmnb_@+~<n|?BN+78dOzCR)ZmVLKhgq=0te=v3G^oQRL7N<sjme3^jPRaLX
z2OX^*OYQ%*^`6P+v@Z@xh6nr}{QkY8dyRdpnb^hygJ1aZrxSBd_`h%rp0T;`!NKsE
z1>g8jY1~fbkZ|bh@k6)&5Sa4*zMQ%3u`}eSV<R15KJ(oU?)%f3^_I5B4~#2zJ$;+3
z!@;kIE#4nuy3We2-LNWn66StPF8prm`t=uXKV8+X+gp#v3U^Ik;oN#;V~>fAj!#pX
zkk_WZzv|wnU;k#*n$w3n>Q5hAy7ZaoqgHqvvE12#Jp04}@%iRcVjr$NH*;5AaZ=~^
z#%}CB!SD`n8}Pla_21WP?$q#?ljkh_X~tLkj-HtMl6RdwzE#(J_WiEU3Wk4;{F6@0
zmnVJ_68s*!$99_|U!}L$)=tjy@c2COwtZr5#)?rH?6_*=lPjGkm9Kp8<|<Jm)f=ug
z&(0nEqWX*T<}bhQIiCAr-sVMXN3UCcbnRwh&Fn5#2@6kp_PgBN_r81n_W`EhtKX%z
zTUWSNBCjn!aon?UJzJ;zRk!6H%L}0H&Qm`;Q8(_Xk?X%awo`p@|L(_*&7ac_Yr9%I
zb^2ga^{P1opM3Bsa=U-}>~`$+Mgv~2U*14A&pI~E{@i%{tnbmq%broc_IU5N8yq@l
z9hn*1rB}ykNRQ2ZTScaKu#Y*atog_}_JjA%x=w7HuuZ&t!P516r62fnZCkWE_F8n;
z;LpxIv)FfFS!d(RH*w+D*3W+T>q8^sqomtsy~f-+iGO_8xfwq%e{A^6x+6Kb+2s3X
zBimDJwyaDIZuZauixD(s^;)^*B5(Jz<WIsiBSQP`_EIxjZ5>!QbhCKihJMc$dyWsj
zHr1%t<!j3#|LduS%x~+&-{t)I{$9Nwvo&~P>C%DYe>(Kq&c?-0yW5L9KMh;K=l7m_
zs?mshdYd|a7P@~S@9fI=cBR&Q@bTyY&n%u^=p9}N#a4ZmYx`gkTRdyls?I3cyx;ww
z+fzu7Jzr*f3(u{dHz@e>+O6H=k8>xOX?^ZkU9$2oo14cxY3I6Z`h59UTUHZq+$CKj
zPe45%f1qD*=*IC+FMslJdhx_%o$D4p)uqd_N7%Yg`o6?v&9B@WDmS3!;-~))+O+jB

literal 0
HcmV?d00001

diff --git a/first version/AccessLocalFunctions.m b/first version/AccessLocalFunctions.m
new file mode 100644
index 0000000..1953044
--- /dev/null
+++ b/first version/AccessLocalFunctions.m	
@@ -0,0 +1,12 @@
+
+
+f = [];
+for i = 1:length(fcns),
+    eval(['f.' char(fcns{i}) ' = fcns{i};']);
+end
+
+%% test function 'FindCentroid'
+M = getappdata(hfig,'M');
+gIX = getappdata(hfig,'gIX');
+C_test = f.FindCentroid(gIX,M);
+figure;imagesc(C_test)
\ No newline at end of file
diff --git a/first version/BasicPlotMaps.m b/first version/BasicPlotMaps.m
new file mode 100644
index 0000000..9196801
--- /dev/null
+++ b/first version/BasicPlotMaps.m	
@@ -0,0 +1,261 @@
+function BasicPlotMaps(cIX,gIX,M,CInfo,photostate,anat_yx,anat_yz,anat_zx) %(cIX,gIX,cIX_0,M_0,CIF,numK,M)
+% M = M_0(cIX_0(cIX),:);
+numK = length(unique(gIX));
+%%
+dataFR = 1;
+figure('Position',[100 50 1350 900]);%,'DeleteFcn',@closefigure_Callback);
+h1 = axes('Position',[0.05, 0.06, 0.5, 0.85]); % left ~subplot
+BasicDrawClusters(h1,M,gIX,dataFR,numK,photostate);
+
+h2 = axes('Position',[0.58, 0.06, 0.40, 0.85]); % right ~subplot
+DrawClustersOnMap_LSh(CInfo,cIX,gIX,numK,anat_yx,anat_yz,anat_zx,'hsv');
+% set(gcf,'PaperUnits', 'inches', 'PaperSize', [12, 6])
+
+end
+
+function BasicDrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,iswrite)
+pos = get(gca,'Position');
+barratio = 0.02;
+numK = double(max(double(numK),double(max(gIX))));
+
+% sort traces by index
+im = M;
+[nLines,nFrames] = size(im);
+[~,I] = sort(gIX);
+im = im(I,:);
+
+%% convert imagesc effect into rgb matrix 'RGB'
+if dataFR, % is drawing cell-response fluorescence
+    cmap = gray(64);
+    im = AutoScaleImage0to1(im); % scaling to min/max 0/1
+    minlim = 0;
+    maxlim = 1;
+else % create blue-white-red map, for regression results
+    cmap = zeros(64,3);
+    cmap(:,1) = [linspace(0,1,32), linspace(1,1,32)];
+    cmap(:,2) = [linspace(0,1,32), linspace(1,0,32)];
+    cmap(:,3) = [linspace(1,1,32), linspace(1,0,32)];
+    minlim = -1; %min(min(im));
+    maxlim = 1; %max(max(im));
+end
+RGB = ImageToRGB(im,cmap,minlim,maxlim); % map image matrix to range of colormap
+
+%% add vertical color code
+if exist('clrmap','var'),
+    if strcmp(clrmap,'jet'),
+        temp = flipud(jet(numK));
+    else % 'hsv'
+        temp = hsv(round(numK*1.1));
+    end
+else % 'hsv'
+    temp = hsv(round(numK*1.1));
+end
+cmap2 = vertcat(temp(1:numK,:),[0,0,0]); % extend colormap to include black
+bwidth = max(round(nFrames/30),1);
+
+idx = gIX(I);
+ix_div = [find(diff(idx));length(idx)];
+
+bars = ones(nLines,bwidth);
+if numK>1,
+    bars(1:ix_div(1),:) = idx(ix_div(1));
+    for i = 2:length(ix_div),
+        % paint color
+        bars(ix_div(i-1)+1:ix_div(i),:) = idx(ix_div(i));
+    end
+end
+im_bars = reshape(cmap2(bars,:),[size(bars),3]);
+% add a white division margin
+div = ones(nLines,round(bwidth/2),3);
+% put 3 parts together
+im = horzcat(RGB,div,im_bars);
+
+%% plot figure
+% h_fig = figure('Position',[100 0 1300 900],'Name',['round ' num2str(i_step)]); %[left, bottom, width, height]
+[s1,s2,s3] = size(im);
+
+hold off;
+set(h1,'Position',[pos(1),pos(2)+pos(4)*barratio*3,pos(3),pos(4)*(1-4*barratio)]); 
+if s1<30,
+    temp = im;
+    im = ones(30,s2,s3);
+    im(1:s1,:,:) = temp;
+end
+image(im);
+set(gca, 'box', 'off')
+
+hold on;
+
+% plot cluster division lines
+if numK>1,
+    for i = 1:length(ix_div),% = numK-1,
+        y = ix_div(i)+0.5;
+        plot([0.5,s2+0.5],[y,y],'k','Linewidth',0.5);
+%         plot([1,s2*kx],[y,y],'k','Linewidth',0.5);
+    end
+end
+
+ylabel(['ROI number: ' num2str(s1)]);
+set(gca,'YTick',[],'XTick',[]);
+set(gcf,'color',[1 1 1]);
+
+colormap gray;
+
+% write in number label
+x = s2*1.003;
+y_last = 0;
+for i = 1:length(ix_div),
+    % avoid label crowding
+    margin = 0.015*s1;
+    y0 = ix_div(i)+0.5;
+    y = max(y_last+margin,y0);
+    if i<length(ix_div),
+        ynext = ix_div(i+1);
+    else
+        ynext = y0+margin*2;
+    end
+    % draw if not squeezed too much away
+    if y < y0+margin*2 && y < 0.5*(y0+ynext), %nLines+stimheight-margin,
+        if exist('iswrite','var') && iswrite,
+            try
+                text(x,y,[num2str(idx(ix_div(i))) ': ' num2str(rankscore(i))],'HorizontalAlignment','Left','VerticalAlignment','Bottom',...
+                    'FontUnits','normalized','FontSize',0.015,'Tag','label');
+            catch
+                disp('rankscore invalid');
+                text(x,y,num2str(idx(ix_div(i))),'HorizontalAlignment','Left','VerticalAlignment','Bottom',...
+                    'FontUnits','normalized','FontSize',0.015,'Tag','label');
+            end
+            
+        else
+            text(x,y,num2str(idx(ix_div(i))),'HorizontalAlignment','Left','VerticalAlignment','Bottom',...
+                'FontUnits','normalized','FontSize',0.015,'Tag','label');
+        end
+        y_last = y;
+    end
+end
+
+if exist('stim','var'),
+    barlength = size(im,2);
+    DrawBars(pos,barratio,nFrames,barlength,stim);
+    
+    if exist('fictive','var'),
+        % Draw fictive
+        axes('Position',[pos(1),pos(2),pos(3),pos(4)*barratio*2]);
+        temp = ones(size(fictive,1),barlength);
+        temp(:,1:length(fictive)) = fictive;
+        imagesc(temp);colormap hot
+        axis off;
+    end
+end
+
+end
+
+%% Internal Functions
+
+function im = AutoScaleImage0to1(im)
+temp = mat2gray(im);
+low_high = stretchlim(temp,[0.005,0.995]);
+temp(temp<low_high(1)) = low_high(1);
+temp(temp>low_high(2)) = low_high(2);
+im = mat2gray(temp);
+end
+
+function RGB = ImageToRGB(im,cmap,minlim,maxlim)
+L = size(cmap,1);
+ix = round(interp1(linspace(minlim,maxlim,L),1:L,im,'linear','extrap'));
+RGB = reshape(cmap(ix,:),[size(ix) 3]); % Make RGB image from scaled.
+end
+
+function DrawBars(pos,barratio,nFrames,barlength,photostate) % horizontal stimulus bar
+% pos = get(gca,'Position'); % [0.05, 0.06, 0.5, 0.85]
+
+% global htop hbottom;
+barheight = 100;
+stimheight = 2*barheight;
+
+m1 = 0.3*ones(barheight,length(photostate));
+m2 = m1; % bottom half
+x = photostate;
+    
+%     2,3,4,5,12,13,14,15,99
+%     3,1,3,2,4 ,10,11,12,
+    
+if max(x)<=3,
+    % 0 = all black; 1 = black/white; 2 = white/black; 3 = all white; (4 = all gray;)
+    % 5 = gray/black; 6 = white/gray; 7 = black/gray; 8 = gray/white.
+    
+    % top (~ projection right)
+    m1(:,x==0 | x==2 | x==5) = 0; % black
+    m1(:,x==4 | x==6 | x==7) = 0.5; % grey
+    m1(:,x==1 | x==3 | x==8) = 1; % white
+    % bottom (~ projection left)
+    m2(:,x==0 | x==1 | x==7) = 0;
+    m2(:,x==4 | x==5 | x==8) = 0.5;
+    m2(:,x==2 | x==3 | x==6) = 1;
+    
+    temp = repmat(vertcat(m1,m2),[1 1 3]); % 3 color layers in dim3
+else
+    %%
+    % 2 = white/white
+    % 3 = black/white
+    % 4 = white/white
+    % 5 = white/black
+    %
+    % 12 = gray
+    % 13 = forward grating (very slow, more for calibration)
+    % 14 = rightward grating
+    % 15 = leftward grating  (* I hope I got right/left correct - if you think it's flipped it probably is)
+    %
+    % 99 = spontaneous
+    % top (~ projection right)
+    
+    m1(:,x==3) = 0; % black
+    m1(:,x==2 | x==4 | x==5) = 1; % white
+    % bottom (~ projection left)
+    m2(:,x==5) = 0;
+    m2(:,x==2 | x==3 | x==4) = 1;
+    
+    temp = repmat(vertcat(m1,m2),[1 1 3]); % 3 color layers in dim3
+    %%
+    temp(:,x==14,1) = 1; % red
+    temp(:,x==13,2) = 1; % green
+    temp(:,x==15,3) = 1; % blue
+    
+    % grey
+    temp(:,x==12,1) = 0.8;
+    temp(:,x==12,2) = 0.8;
+    temp(:,x==12,3) = 0.8;
+    
+    % yellow
+    temp(:,x==99,1) = 1;
+    temp(:,x==99,2) = 1;
+    temp(:,x==99,3) = 0.6;
+end
+
+n = ceil(nFrames/length(photostate)); % tile, to size or bigger
+temp = repmat(temp,1,n);
+stimbar = ones(stimheight,barlength,3);
+stimbar(:,1:nFrames,:) = temp(:,1:nFrames,:); % crop to size
+
+% if isempty(htop),
+axes('Position',[pos(1),pos(2)+pos(4)*(1-barratio),pos(3),pos(4)*barratio]);
+% elseif ~ishandle(htop),
+%     htop = axes('Position',[pos(1),pos(2)+pos(4)*(1-barratio),pos(3),pos(4)*barratio]);
+% else axes(htop)
+% end
+image(stimbar);axis off;hold on;
+% plot top border
+plot([1,length(photostate)],[1,1],'k','Linewidth',0.5);
+
+% if isempty(hbottom),
+axes('Position',[pos(1),pos(2)+pos(4)*barratio*2,pos(3),pos(4)*barratio]);
+% elseif ~ishandle(hbottom),
+%     hbottom = axes('Position',[pos(1),pos(2),pos(3),pos(4)*barratio]);
+% else axes(hbottom)
+% end
+
+image(stimbar);axis off;hold on;
+% plot bottom border
+plot([1,length(photostate)],[stimheight,stimheight],'k','Linewidth',0.5);
+%%
+end
diff --git a/Conv_test.m b/first version/Conv_test.m
similarity index 100%
rename from Conv_test.m
rename to first version/Conv_test.m
diff --git a/DA_2_Solution.m b/first version/DA_2_Solution.m
similarity index 100%
rename from DA_2_Solution.m
rename to first version/DA_2_Solution.m
diff --git a/first version/DrawClusters.m b/first version/DrawClusters.m
new file mode 100644
index 0000000..964da88
--- /dev/null
+++ b/first version/DrawClusters.m	
@@ -0,0 +1,210 @@
+function DrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,iswrite)
+pos = get(gca,'Position');
+barratio = 0.03;
+
+%% Prepare cluster data
+% down-sample
+displaymax = 1000;
+numcell = size(M,1);
+if numcell > displaymax,
+    skip = round(numcell/displaymax);
+    M = M(1:skip:end,:);
+    gIX = gIX(1:skip:end,:);
+end
+
+numK = double(max(double(numK),double(max(gIX))));
+
+% sort traces by index
+im = M;
+[nLines,nFrames] = size(im);
+[~,I] = sort(gIX);
+im = im(I,:);
+
+%% convert imagesc effect into rgb matrix 'RGB'
+if dataFR, % is drawing cell-response fluorescence
+    cmap = gray(64);
+    im = AutoScaleImage0to1(im); % scaling to min/max 0/1
+    minlim = 0;
+    maxlim = 1;
+else % create blue-white-red map, for regression results
+    cmap = zeros(64,3);
+    cmap(:,1) = [linspace(0,1,32), linspace(1,1,32)];
+    cmap(:,2) = [linspace(0,1,32), linspace(1,0,32)];
+    cmap(:,3) = [linspace(1,1,32), linspace(1,0,32)];
+    minlim = -1; %min(min(im));
+    maxlim = 1; %max(max(im));
+end
+RGB = ImageToRGB(im,cmap,minlim,maxlim); % map image matrix to range of colormap
+
+%% add vertical color code
+if exist('clrmap','var'),
+    if strcmp(clrmap,'jet'),
+        temp = flipud(jet(numK));
+    else % 'hsv'
+        temp = hsv(round(numK*1.1));
+    end
+else % 'hsv'
+    temp = hsv(round(numK*1.1));
+end
+cmap2 = vertcat(temp(1:numK,:),[0,0,0]); % extend colormap to include black
+bwidth = max(round(nFrames/30),1);
+
+idx = gIX(I);
+ix_div = [find(diff(idx));length(idx)];
+
+bars = ones(nLines,bwidth);
+if numK>1,
+    bars(1:ix_div(1),:) = idx(ix_div(1));
+    for i = 2:length(ix_div),
+        % paint color
+        bars(ix_div(i-1)+1:ix_div(i),:) = idx(ix_div(i));
+    end
+end
+im_bars = reshape(cmap2(bars,:),[size(bars),3]);
+% add a white division margin
+div = ones(nLines,round(bwidth/2),3);
+% put 3 parts together
+im = horzcat(RGB,div,im_bars);
+
+%% plot figure
+
+[s1,s2,s3] = size(im);
+
+hold off;
+set(h1,'Position',[pos(1),pos(2)+pos(4)*barratio*3,pos(3),pos(4)*(1-4*barratio)]); 
+if s1<30,
+    temp = im;
+    im = ones(30,s2,s3);
+    im(1:s1,:,:) = temp;
+end
+image(im);
+set(gca, 'box', 'off')
+
+hold on;
+
+% plot cluster division lines
+plot([0.5,s2+0.5],[0.5,0.5],'k','Linewidth',0.5);
+if numK>1,
+    for i = 1:length(ix_div),% = numK-1,
+        y = ix_div(i)+0.5;
+        plot([0.5,s2+0.5],[y,y],'k','Linewidth',0.5);
+    end
+end
+
+ylabel(['Number of cells: ' num2str(numcell)]);
+set(gca,'YTick',[],'XTick',[]);
+set(gcf,'color',[1 1 1]);
+
+colormap gray;
+
+% write in number label
+x = s2*1.003;
+y_last = 0;
+for i = 1:length(ix_div),
+    % avoid label crowding
+    margin = 0.015*s1;
+    y0 = ix_div(i)+0.5;
+    y = max(y_last+margin,y0);
+    if i<length(ix_div),
+        ynext = ix_div(i+1);
+    else
+        ynext = y0+margin*2;
+    end
+    % draw if not squeezed too much away
+    if y < y0+margin*2 && y < 0.5*(y0+ynext), %nLines+stimheight-margin,
+        if exist('iswrite','var') && iswrite,
+            try
+                text(x,y,[num2str(idx(ix_div(i))) ': ' num2str(rankscore(i))],'HorizontalAlignment','Left','VerticalAlignment','Bottom',...
+                    'FontUnits','normalized','FontSize',0.015,'Tag','label');
+            catch
+                disp('rankscore invalid');
+                text(x,y,num2str(idx(ix_div(i))),'HorizontalAlignment','Left','VerticalAlignment','Bottom',...
+                    'FontUnits','normalized','FontSize',0.015,'Tag','label');
+            end
+            
+        else
+            text(x,y,num2str(idx(ix_div(i))),'HorizontalAlignment','Left','VerticalAlignment','Bottom',...
+                'FontUnits','normalized','FontSize',0.015,'Tag','label');
+        end
+        y_last = y;
+    end
+end
+
+%% Draw stim bars
+halfbarheight = 100;
+stimbar_ = GetStimBar(halfbarheight,stim);
+
+% pad with margin on right to fit combo-image size
+barlength = size(im,2);
+stimbar = ones(halfbarheight*2,barlength,3);
+stimbar(:,1:nFrames,:) = stimbar_(:,1:nFrames,:); % crop to size
+
+% top axis
+axes('Position',[pos(1),pos(2)+pos(4)*(1-barratio),pos(3),pos(4)*barratio]);
+image(stimbar);axis off;hold on;
+plot([1,length(stim)],[1,1],'k','Linewidth',0.5); % plot top border
+plot([0.5,0.5],[0,size(stimbar,1)],'k','Linewidth',0.5); % plot left/right border
+plot([length(stim),length(stim)],[0,size(stimbar,1)],'k','Linewidth',0.5);
+
+% bottom axis
+axes('Position',[pos(1),pos(2)+pos(4)*barratio*2,pos(3),pos(4)*barratio]);
+image(stimbar);axis off;hold on;
+plot([1,length(stim)],[size(stimbar,1),size(stimbar,1)],'k','Linewidth',0.5); % plot bottom border
+plot([0.5,0.5],[0,size(stimbar,1)],'k','Linewidth',0.5); % plot left/right border
+plot([length(stim),length(stim)],[0,size(stimbar,1)],'k','Linewidth',0.5);
+
+%% Draw fictive
+axes('Position',[pos(1),pos(2),pos(3),pos(4)*barratio*2]);
+
+temp = ones(size(fictive,1),barlength);
+temp(:,1:length(fictive)) = fictive;
+imagesc(temp);colormap hot
+set(gca,'YTick',[],'XTick',[]);
+set(gcf,'color',[1 1 1]);
+set(gca, 'box', 'off')
+hold on;axis ij;
+
+% plot division lines
+for i = 0:3,
+    y = i+0.5;
+    plot([0.5,length(fictive)+0.5],[y,y],'w','Linewidth',0.5);
+end
+% labels
+names = {'Left','Right','Forward','Raw L','Raw R'};
+x = -s2*0.05;
+for i = 1:5,
+    y = i;
+    text(x,y,names{i},'Fontsize',7);
+end
+
+%% horizontal axis: approximate time length, assuming 2 fps (real is 1.97 for now)
+set(gca,'XColor',[1 1 1]);
+if s2/2/60<5, % <5 min
+    l = xlabel(['Length of display: ~ ' num2str(s2/2) 'sec'],'Color',[0 0 0]);
+else
+    l = xlabel(['Length of display: ~ ' num2str(s2/2/60) 'min'],'Color',[0 0 0]);
+end
+
+%% Draw fish icon:
+if exist('fishpic.jpg','file')==2,
+    fishpic = imread('fishpic.jpg');
+    axes('Position',[0.015, 0.095, 0.03, 0.02]);
+    image(fishpic); axis off; axis equal
+end
+end
+
+%% Internal Functions
+
+function im = AutoScaleImage0to1(im)
+temp = mat2gray(im);
+low_high = stretchlim(temp,[0.005,0.995]);
+temp(temp<low_high(1)) = low_high(1);
+temp(temp>low_high(2)) = low_high(2);
+im = mat2gray(temp);
+end
+
+function RGB = ImageToRGB(im,cmap,minlim,maxlim)
+L = size(cmap,1);
+ix = round(interp1(linspace(minlim,maxlim,L),1:L,im,'linear','extrap'));
+RGB = reshape(cmap(ix,:),[size(ix) 3]); % Make RGB image from scaled.
+end
diff --git a/first version/DrawClustersOnMap_LSh.m b/first version/DrawClustersOnMap_LSh.m
new file mode 100644
index 0000000..5bbeb48
--- /dev/null
+++ b/first version/DrawClustersOnMap_LSh.m	
@@ -0,0 +1,137 @@
+function  [tot_image, dim_totimage] = DrawClustersOnMap_LSh(CInfo,cIX,gIX,numK,anat_yx,anat_yz,anat_zx,clrmap,full)
+%% formatting
+[s1,s2] = size(cIX);
+if s2>s1,
+    cIX = cIX';
+end
+[s1,s2] = size(gIX);
+if s2>s1,
+    gIX = gIX';
+end
+
+% down-sample
+if ~exist('full','var'),
+    displaymax = 8000;
+    if length(cIX) > displaymax,
+        skip = round(length(cIX)/displaymax);
+        cIX = cIX(1:skip:end,:);
+        gIX = gIX(1:skip:end,:);
+    end
+end
+
+% get numK
+if exist('numK','var'),
+    numK = double(max(numK,max(gIX)));
+else
+    numK = double(max(gIX));
+end
+
+if strcmp(clrmap,'jet'),
+    temp = flipud(jet(numK));
+else % 'hsv'
+    temp = hsv(round(numK*1.1));
+end
+cmap = temp(1:numK,:); % extend colormap to include black
+
+anat_YX = anat_yx/4;
+anat_YZ = anat_yz/4;
+anat_ZX = anat_zx/4;
+anat_ZX = flipud(anat_ZX);
+dimv_yx = size(anat_YX);
+dimv_yz = size(anat_YZ);
+dimv_zx = size(anat_ZX);
+
+k_zres = 20;
+anat_yz2=zeros(dimv_yz(1),dimv_yz(2)*k_zres,3);
+anat_zx2=zeros(dimv_zx(1)*k_zres,dimv_zx(2),3);
+dim_totimage = [dimv_yx(1)+dimv_zx(1)*k_zres+10,dimv_yx(2)+dimv_yz(2)*k_zres+10,3];
+tot_image=ones(dim_totimage);
+% tot_image=zeros(dim_totimage);
+% tot_image(:,dimv_yx(2)+(1:10),:)=1;
+
+% find index manipulation vector to darw circle
+circle=makeDisk2(7,15); % make mask of filled circle % (7,15)
+mask = zeros(dimv_yx(1),dimv_yx(2));
+mask(1:15,1:15) = circle;
+ix = find(mask);
+cix = sub2ind([dimv_yx(1),dimv_yx(2)],8,8);% 8
+circle_inds = ix - cix;
+
+yzplane_inds = -5:5;
+zxplane_inds = -5*dimv_zx(1):dimv_zx(1):5*dimv_zx(1);
+
+weight = 0.3-min(length(cIX)/1000/100,0.1);
+for j=1:length(cIX)
+    if ~isempty(CInfo(cIX(j)).center),
+        %% Y-X
+        %     cinds = sub2ind([dim_y,dim_x],cell_info(cellsIX(j)).center(1),cell_info(cellsIX(j)).center(2));
+        cinds=(CInfo(cIX(j)).center(2)-1)*dimv_yx(1)+CInfo(cIX(j)).center(1); % faster equivalent, lin px idx
+        labelinds=find((cinds+circle_inds)>0 & (cinds+circle_inds)<=dimv_yx(1)*dimv_yx(2)); % within bounds
+        ix = gIX(j);%find(U==gIX(j));U = 1:numK;
+        ixs = cinds+circle_inds(labelinds);
+        anat_YX(ixs) = cmap(ix,1)*weight + anat_YX(ixs)*(1-weight); % R
+        ixs = cinds+circle_inds(labelinds)+dimv_yx(1)*dimv_yx(2);
+        anat_YX(ixs) = cmap(ix,2)*weight + anat_YX(ixs)*(1-weight); % G
+        ixs = cinds+circle_inds(labelinds)+dimv_yx(1)*dimv_yx(2)*2;
+        anat_YX(ixs) = cmap(ix,3)*weight + anat_YX(ixs)*(1-weight); % B
+        
+        % Y-Z
+        zweight = weight/2;
+        %     cinds = sub2ind([dim_y,dim_z],cell_info(cellsIX(j)).center(1),cell_info(cellsIX(j)).slice);
+        cinds=(CInfo(cIX(j)).slice-1)*dimv_yz(1)+CInfo(cIX(j)).center(1);
+        labelinds=find((cinds+yzplane_inds)>0 & (cinds+yzplane_inds)<=dimv_yz(1)*dimv_yz(2));
+        ixs = cinds+yzplane_inds(labelinds);
+        anat_YZ(ixs) = cmap(ix,1)*zweight + anat_YZ(ixs)*(1-zweight);
+        ixs = cinds+yzplane_inds(labelinds)+dimv_yz(1)*dimv_yz(2);
+        anat_YZ(ixs) = cmap(ix,2)*zweight + anat_YZ(ixs)*(1-zweight);
+        ixs = cinds+yzplane_inds(labelinds)+dimv_yz(1)*dimv_yz(2)*2;
+        anat_YZ(ixs) = cmap(ix,3)*zweight + anat_YZ(ixs)*(1-zweight);
+                
+        % Z-X
+        zweight = weight/2;
+        %     cinds = sub2ind([dim_y,dim_z],cell_info(cellsIX(j)).center(1),cell_info(cellsIX(j)).slice);
+        cinds=(CInfo(cIX(j)).center(2)-1)*dimv_zx(1) +(CInfo(cIX(j)).slice);
+        labelinds=find((cinds+zxplane_inds)>0 & (cinds+zxplane_inds)<=dimv_zx(1)*dimv_zx(2));
+        ixs = cinds+zxplane_inds(labelinds);
+        anat_ZX(ixs) = cmap(ix,1)*zweight + anat_ZX(ixs)*(1-zweight);
+        ixs = cinds+zxplane_inds(labelinds)+dimv_zx(1)*dimv_zx(2);
+        anat_ZX(ixs) = cmap(ix,2)*zweight + anat_ZX(ixs)*(1-zweight);
+        ixs = cinds+zxplane_inds(labelinds)+dimv_zx(1)*dimv_zx(2)*2;
+        anat_ZX(ixs) = cmap(ix,3)*zweight + anat_ZX(ixs)*(1-zweight);
+    end
+end
+% rescale (low-res) z dimension
+for k=1:3
+    anat_yz2(:,:,1) = imresize(anat_YZ(:,:,1), [dimv_yz(1), dimv_yz(2)*k_zres],'nearest');
+    anat_yz2(:,:,2) = imresize(anat_YZ(:,:,2), [dimv_yz(1), dimv_yz(2)*k_zres],'nearest');
+    anat_yz2(:,:,3) = imresize(anat_YZ(:,:,3), [dimv_yz(1), dimv_yz(2)*k_zres],'nearest');
+    
+    anat_zx2(:,:,1) = imresize(anat_ZX(:,:,1), [dimv_zx(1)*k_zres, dimv_zx(2)],'nearest');
+    anat_zx2(:,:,2) = imresize(anat_ZX(:,:,2), [dimv_zx(1)*k_zres, dimv_zx(2)],'nearest');
+    anat_zx2(:,:,3) = imresize(anat_ZX(:,:,3), [dimv_zx(1)*k_zres, dimv_zx(2)],'nearest');
+end
+
+tot_image(dimv_zx(1)*k_zres+11:end,1:dimv_yx(2),:) = anat_YX;
+tot_image(dimv_zx(1)*k_zres+11:end,dimv_yx(2)+11:end,:) = anat_yz2;
+tot_image(1:dimv_zx(1)*k_zres,1:dimv_zx(2),:) = flipud(anat_zx2);
+
+tot_image(tot_image(:)>1) = 1;
+tot_image(tot_image(:)<0) = 0;
+
+image(tot_image);
+axis image;axis off
+
+end
+
+function out = makeDisk2(radius, dim)
+center=floor(dim/2)+1;
+out=zeros(dim);
+for x=1:dim
+    for y=1:dim
+        if norm([x,y]-[center,center])<=radius
+            out(x,y)=1;
+        end
+    end
+end
+
+end
\ No newline at end of file
diff --git a/FFT_test.m b/first version/FFT_test.m
similarity index 100%
rename from FFT_test.m
rename to first version/FFT_test.m
diff --git a/first version/GUI_FishExplorer.m b/first version/GUI_FishExplorer.m
new file mode 100644
index 0000000..a417875
--- /dev/null
+++ b/first version/GUI_FishExplorer.m	
@@ -0,0 +1,3446 @@
+%%% Interactive app for exploratory analysis of calcium imaging data 
+% (with stimulus, behavior, and anatomy)
+
+% Input calcium data: 1 trace per cell/ROI, ~50,000 cells per fish
+% load collection of cells from multiple fish, or load full data of single fish individually
+% main outputs: GUI plots, clusters saved into .mat, export variables to MATLAB workspace
+
+% Tip: to see the structure of this code, use 'Ctrl' + 'm' + '=' to collapse all cells.
+% UI controls are organized by tabs and then by rows, instructions and 
+% comments are where they are constructed ('User Interface:' -> function hfig... ->)
+% General internal functions are at the end, some specialized .m functions are outside.
+
+% Written in Matlab R2014b running on Windows 7.
+
+% - Xiuye Chen (xiuyechen@gmail.com), Engert Lab, Spring 2015
+
+
+% To start, run the following in separate script or command window:
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+% clear all;close all;clc
+% global VAR;
+% load('VAR_current.mat','VAR');
+% load('CONSTs_current.mat','CONSTs');
+% hfig = GUI_FishExplorer(CONSTs,VAR);
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+% Full datasets are stored as 'CONST_F?.mat' (? = fish number), to load from the GUI
+
+%% User Interface:
+% function hfig = GUI_FishExplorer(CONSTs,VAR,CONST,i_fish)
+function [hfig,fcns] = GUI_FishExplorer(data_dir,CONSTs,VAR,flag_script,var_script)
+if exist('flag_script','var')
+    if ~exist('var_script','var')
+        var_script={};
+    end
+    runscript(flag_script,var_script);
+    return
+end
+
+%% Make figure
+scrn = get(0,'Screensize');
+hfig = figure('Position',[scrn(3)*0.1 scrn(4)*0.05 scrn(3)*0.85 scrn(4)*0.86],...% [50 100 1700 900]
+    'Name','GUI_LSh','DeleteFcn',@closefigure_Callback);
+hold off; axis off
+
+%% Folder setup
+% directory for full fish data (.mat)
+setappdata(hfig,'data_dir',data_dir);
+
+% copy of VAR files will be saved into this subfolder:
+currentfolder = pwd;
+arcmatfolder = [currentfolder '\arc mat'];
+if ~exist(arcmatfolder, 'dir')
+  mkdir(arcmatfolder);
+end
+setappdata(hfig,'arcmatfolder',arcmatfolder);
+
+%% Pass external variables into appdata (stored with main figure handle)
+setappdata(hfig,'CONSTs',CONSTs);
+setappdata(hfig,'VAR',VAR);
+nFish = length(CONSTs);
+
+% parameters / constants
+setappdata(hfig,'z_res',19.7); % resoltion in z, um per slice
+% fpsec = 1.97; % hard-coded in ext function 'GetStimRegressor.m'
+% approx fpsec of 2 used in ext function 'DrawCluster.m'
+
+% cache 
+bC = []; % Cache for going b-ack (bad abbr)
+fC = []; % Cache for going f-orward
+bC.cIX = cell(1,1);
+bC.gIX = cell(1,1);
+bC.numK = cell(1,1);
+fC.cIX = cell(1,1);
+fC.gIX = cell(1,1);
+fC.numK = cell(1,1);
+setappdata(hfig,'bCache',bC);
+setappdata(hfig,'fCache',fC);
+
+% initialization
+i_fish = 1;
+QuickUpdateFish(hfig,i_fish,'init');
+
+%% Initialize internal params into appdata
+% fish protocol sets (different sets have different parameters)
+M_fish_set = [1, 1, 1, 1, 1, 1, 1, 2, 1]; % M = Matrix
+setappdata(hfig,'M_fish_set',M_fish_set);
+
+% thresholds
+thres_merge = 0.9;
+thres_split = 0.7;
+thres_reg = 0.7;
+thres_size = 10;
+setappdata(hfig,'thres_merge',thres_merge);
+setappdata(hfig,'thres_split',thres_split); % for function 'pushbutton_iter_split'
+setappdata(hfig,'thres_reg',thres_reg); % regression threshold, ~correlation coeff
+setappdata(hfig,'thres_size',thres_size); % min size for clusters
+
+% variables 
+% (not sure all these need to be initialized, probably not complete either)
+setappdata(hfig,'clrmap','hsv');
+setappdata(hfig,'opID',0);
+setappdata(hfig,'rankID',0); 
+setappdata(hfig,'rankscore',[]);
+setappdata(hfig,'isCentroid',0);
+setappdata(hfig,'isWkmeans',1); % in autoclustering, with/without kmeans
+setappdata(hfig,'isflipstim',0); % flag for flip updown
+setappdata(hfig,'regchoice',{1,1}); % regressor choice; first cell,1=stim,2=motor,3=centroid
+setappdata(hfig,'isfullfish',0); % no if QuickUpdateFish, yes if LoadFullFish
+setappdata(hfig,'isPlotCorrHist',0); % option for regression
+setappdata(hfig,'dataFR',1); % left plot data type, 1 for fluo trace and 0 for regression result
+setappdata(hfig,'isPlotReg',1); % plot regressor when selecting it
+setappdata(hfig,'hierinplace',1); % hier. partitioning, no reordering
+
+
+Class = getappdata(hfig,'Class');
+classID = numel(Class);
+setappdata(hfig,'classID',classID);
+setappdata(hfig,'classheader','Test:');
+setappdata(hfig,'newclassname','');
+
+Cluster = getappdata(hfig,'Cluster');
+clusID = 1; %numel(Cluster);
+setappdata(hfig,'clusID',clusID);
+setappdata(hfig,'clusheader','Test:');
+setappdata(hfig,'newclusname','');
+
+setappdata(hfig,'cIX',Cluster(clusID).cIX); % cell-IndeX
+setappdata(hfig,'gIX',Cluster(clusID).gIX); % group-IndeX
+
+%% Create UI controls
+set(gcf,'DefaultUicontrolUnits','normalized');
+set(gcf,'defaultUicontrolBackgroundColor',[1 1 1]);
+
+% tab group setup
+tgroup = uitabgroup('Parent', hfig, 'Position', [0.05,0.88,0.91,0.12]);
+numtabs = 5;
+tab = cell(1,numtabs);
+M_names = {'General','Operations','Regression','Clustering etc.','Saved Clusters'};
+for i = 1:numtabs,
+    tab{i} = uitab('Parent', tgroup, 'BackgroundColor', [1,1,1], 'Title', M_names{i});
+end
+
+% grid setup, to help align display elements
+rheight = 0.2;
+yrow = 0.7:-0.33:0;%0.97:-0.03:0.88;
+bwidth = 0.03;
+grid = 0:bwidth+0.001:1;
+
+% various UI element handles ('global' easier than passing around..)
+global hback hfwd hclassname hclassmenu hclusgroupmenu hclusmenu hclusname...
+    hdatamenu hopID hcentroid hdataFR hloadfish hfishnum hstimreg hmotorreg...
+    hcentroidreg;
+
+%% UI ----- tab one ----- (General)
+i_tab = 1;
+
+%% UI row 1: File 
+i_row = 1;
+i = 1;n = 0;
+
+i=i+n;
+n=2; % saves 'VAR' to workspace. 'VAR' contains clustering indices (Class and Clusgroup) for all fish
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Quick save',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_save_Callback);
+
+i=i+n;
+n=2; % saves 'VAR' both to workspace and to 'VAR_current.mat' and to arc folder
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Save .mat',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_savemat_Callback);
+
+i=i+n;
+n=2; % plots selected cells on anatomy z-stack, display and save tiff stack in current directory
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Save Zstack',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_writeZstack_Callback);
+
+i=i+n;
+n=2; % make new figure without the GUI components, can save manually from default menu
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Popup plot',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_popupplot_Callback);
+
+i=i+n;
+n=3; % export main working variables to workspace, can customize!
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Export to workspace',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_exporttoworkspace_Callback);
+
+%% UI row 2: Load 
+i_row = 2;
+i = 1;n = 0;
+
+i=i+n;
+n=2; % this design is underused now... Quick-load only depends on CONSTs,
+% which is a minimum collection of clusters from all fish, so you can load
+% the program without full single-fish data. eventually can use this  
+% platform to do things across fish (like after anatomical alignment).
+uicontrol('Parent',tab{i_tab},'Style','text','String','Quick-load fish:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1; % loads 'CONSTs_current.mat' from current directory
+temp = {}; for j = 1:nFish, temp = [temp,{num2str(j)}];end
+hfishnum = uicontrol('Parent',tab{i_tab},'Style','popupmenu','String',temp,...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@popup_fishmenu_Callback);
+
+i=i+n;
+n=2; % loads full single-fish data from CONST_F?.mat
+uicontrol('Parent',tab{i_tab},'Style','text','String','Load full fish:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=2; % these CONST_F?.mat are now saved as uncompressed version 6 .mat, loads faster
+temp = {}; for j = 1:nFish, temp = [temp,{num2str(j)}];end
+temp = [{'(choose)'},temp];
+hloadfish = uicontrol('Parent',tab{i_tab},'Style','popupmenu','String',temp,...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@popup_loadfullfishmenu_Callback);
+
+i=i+n;
+n=2; % load different presentations (average, all reps etc), pre-stored in CONST(s)
+uicontrol('Parent',tab{i_tab},'Style','text','String','Data type:',... 
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=3; 
+hdatamenu = uicontrol('Parent',tab{i_tab},'Style','popupmenu','String',CONSTs{i_fish}.datanames,...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@popup_datamenu_Callback);
+
+i=i+n;
+n=4; % only centroids (~mean) of clusters shown on left-side plot, the rest is unchanged
+hcentroid = uicontrol('Parent',tab{i_tab},'Style','checkbox','String','Display centroids of clusters',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@checkbox_showcentroids_Callback);
+
+%% UI ----- tab two ----- (Operations)
+i_tab = 2;
+
+%% UI row 1: Range
+i_row = 1;
+i = 1;n = 0;
+
+i=i+n;
+n=2; % saves up to 20 steps backwards (datatype/datamenu change does not count)
+hback = uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Back',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_back_Callback);
+
+i=i+n;
+n=2; % same, 20 steps forward if applicable
+hfwd = uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Forward',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_forward_Callback);
+
+i=i+n;
+n=3; % Choose range of clusters to keep. format: e.g. '1:2,4-6,8:end'
+uicontrol('Parent',tab{i_tab},'Style','text','String','Choose cluster range:',... 
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1; 
+uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_choose_range_Callback);
+
+i=i+n;
+n=2; % Choose range of clusters to exclude. format: e.g. '1:2,4-6,8:end'
+uicontrol('Parent',tab{i_tab},'Style','text','String','Exclude:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_exclude_range_Callback);
+
+i=i+n;
+n=1; % Choose range of clusters to fuse/combine into single cluster. format: e.g. '1:2,4-6,8:end'
+uicontrol('Parent',tab{i_tab},'Style','text','String','Fuse:',... % (eg 1:2,3-5)
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_fuse_range_Callback);
+
+i=i+n;
+n=2; % can choose range in time, but can't see time axes to decide... resets when choosing datamenu
+uicontrol('Parent',tab{i_tab},'Style','text','String','Time-range:',... % (eg 1:2,3-5)
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_t_range_Callback);
+
+%% UI row 2: Operations
+i_row = 2;
+i = 1;n = 0;
+
+i=i+n;
+n=2; % operates between the current cell selection and the next (in this order). 
+uicontrol('Parent',tab{i_tab},'Style','text','String','Set operations:',... 
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n; % 'setdiff' is current minus next, 'rev setdiff' is next minus current. 
+n=2; % smartUnion = SmartUnique, cells belonging to 2 clusters goes to the more correlated one
+menu = {'(choose)','union','intersect','setdiff','rev setdiff','setxor','smartUnion'};
+hopID = uicontrol('Parent',tab{i_tab},'Style','popupmenu','String',menu,'Value',1,...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@popup_operations_Callback});
+
+i=i+n;
+n=2; % rank clusters based on various criteria (to choose)
+uicontrol('Parent',tab{i_tab},'Style','text','String','Rank by:',... 
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n; % 'hier' is the same as default (used after every k-means);'stim-lock' uses std across reps;
+n=2; % motor stuff uses the best alignment (by cross-correlation) with the fictive trace;
+% L+R is average of L & R; stim-motor is combines 'stim-lock' w 'motor' with arbituary weighting.
+menu = {'(choose)','hier.','size','stim-lock','corr',...
+    'motor','L motor','R motor','L+R motor','stim-motor'}; 
+uicontrol('Parent',tab{i_tab},'Style','popupmenu','String',menu,'Value',1,...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@popup_ranking_Callback});
+
+i=i+n;
+n=3; % cluster indices will rank from 1 to number of clusters
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Sqeeze clusters',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_sqeeze_Callback);
+
+i=i+n;
+n=3; % flip the sequenc of clusters, first becomes last
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Flip up-down',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_flipud_Callback);
+
+i=i+n;
+n=3; % switch between 2 colormaps now, jet and a cropped version of hsv (so not all circular)
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Switch colormap',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_clrmap_Callback});
+
+%% UI row 3: Anatomy
+i_row = 3;
+i = 1;n = 0;
+
+i=i+n;
+n=4; % Draw a polygon on anatomy maps to select the cells within those boundaries
+uicontrol('Parent',tab{i_tab},'Style','text','String','Draw on anatomy map to crop:',... 
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n; % Draw on the yx-view (main view) 
+n=2; % Click to make new vertex, double click to connect to first vertex, 
+% then optionally drag vertices to reposition, and finally double click again to set
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Draw yx',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_polygon_yx_Callback);
+
+i=i+n;
+n=2; % Draw on yz-view (side projection), same as above
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Draw yz',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_polygon_yz_Callback);
+
+i=i+n;
+n=2; % Draw on zx-view (front projection), same as above
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Draw zx',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_polygon_zx_Callback);
+
+i=i+n;
+n=5; 
+uicontrol('Parent',tab{i_tab},'Style','text','String','Select all cells within boundaries:',... 
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=2; % selects ALL cells contained in dataset that are within the convex shape defined by current cells
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Convex hull',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_withinConvexHull_Callback);
+
+%% UI ----- tab three ----- (Regression)
+i_tab = 3;
+
+%% UI row 1: regressor
+i_row = 1; % Step 1:
+i = 1;n = 0; % Choose one type of regressor here, choice highlighted in yellow
+
+i=i+n;
+n=2; % stimulus regressors, go to 'GetStimRegressor.m' to add/update
+uicontrol('Parent',tab{i_tab},'Style','text','String','Stim reg.:',... 
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=3; % (updated when loading fish)
+menu = {'(choose)',''};
+hstimreg = uicontrol('Parent',tab{i_tab},'Style','popupmenu','String',menu,'Value',1,...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@popup_getstimreg_Callback);
+
+i=i+n; % motor regressors from fictive, not yet convolved/adjusted for time lag
+n=2; % go to 'GetMotorRegressor.m' to add/update
+uicontrol('Parent',tab{i_tab},'Style','text','String','Motor reg.:',... 
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=2; % (unlike stim regressors, names hardcoded, not importet from regressor...)
+menu = {'(choose)','right swims','left swims','forward swims','raw right','raw left','raw average'};
+hmotorreg = uicontrol('Parent',tab{i_tab},'Style','popupmenu','String',menu,'Value',1,...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@popup_getmotorreg_Callback);
+
+i=i+n;
+n=2; % if checked, plot regressor during selection, together with stim and motor
+uicontrol('Parent',tab{i_tab},'Style','checkbox','String','Plot regressor',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],'Value',1,...
+    'Callback',@checkbox_popupplotreg_Callback);
+
+i=i+n+1;
+n=4; % Get centroid (~mean) of selected cluster as regressor
+uicontrol('Parent',tab{i_tab},'Style','text','String','Regressor from centroid #:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1;
+hcentroidreg = uicontrol('Parent',tab{i_tab},'Style','edit','String',num2str(1),...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_ctrdID_as_reg_Callback);
+
+i=i+n;
+n=2; % too complicated... chomp up centroid, then order as if stimulus were left/right inverted
+uicontrol('Parent',tab{i_tab},'Style','checkbox','String','Flip stim',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@checkbox_flipstim_Callback);
+
+%% UI row 2: regression
+i_row = 2; % Step 2:
+i = 1;n = 0; % Choose regression, using the regressor chosen above, search in full dataset
+
+i=i+n;
+n=3;
+uicontrol('Parent',tab{i_tab},'Style','text','String','Choose regression ->',... 
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=2; % do regression, show all cells with correlation coeff (with regressor) above threshold
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Corr. threshold:',... 
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_thres_regression_Callback);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit','String',num2str(thres_reg),...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_regthres_Callback);
+
+i=i+n;
+n=2; % optionally plot histogram of correlation values for all cells in dataset, visualize cut-off
+hdataFR = uicontrol('Parent',tab{i_tab},'Style','checkbox','String','Plot corr. hist',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@checkbox_plotcorrhist_Callback);
+
+i=i+n;
+n=3; % probably not so useful. Show top n cells of highest correlation coeff with regressor
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','top corr., number limit:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_topnum_regression_Callback});
+
+i=i+n;
+n=1; % specify n for above
+uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_topCorrNumber_Callback);
+
+i=i+n+1; % more automatic, do a regression with every centroid, then combine (with 'SmartUnique'
+n=4; % i.e. overlapping cells are assigned to the cluster with which the correlation is the highest)
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Regression with all centroids',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_AllCentroidRegression_Callback});
+
+i=i+n; % this is a remnant button from a failed experiment, idea was to iterate the regression process
+n=2; % until the cluster converges, but most of the time it doesn't...
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','iter.reg','Enable','off',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_IterCentroidRegression_Callback});
+
+%% UI row 3: (TBD)
+i_row = 3; % old idea: display correlation values of multiple regressors (instead of fluo. trace)
+i = 1;n = 0; % and then can cluster that and further manipulate... 
+
+i=i+n;
+n=4; % did not implement combination of regressors in this version... but display option is coded (dataFR==0)
+uicontrol('Parent',tab{i_tab},'Style','text','String','regressor combos??(TBD)',... 
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=4; % also data storage is coded, 'M' could be loaded from M_0_fluo or M_0_reg (storing reg corr values)
+hdataFR = uicontrol('Parent',tab{i_tab},'Style','checkbox','String','Display fluo./regression',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@checkbox_dataFR_Callback);
+
+%% UI ----- tab four ----- (Clustering etc.)
+i_tab = 4;
+
+%% UI row 1: k-means
+i_row = 1;
+i = 1;n = 0;
+
+i=i+n;
+n=2; % k-means clustering
+uicontrol('Parent',tab{i_tab},'Style','text','String','k-means, k =',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_kmeans_Callback);
+
+i=i+n;
+n=4; % anatomy is added to the fluo trace as new dimensions, and (arbituarily) weighted strongly
+uicontrol('Parent',tab{i_tab},'Style','text','String','k-means with anatomy, k =',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_kmeans2_Callback);
+
+i=i+n; % trying to use Silhouette to evaluate cluster quality, find peak to determine optimal k,
+n=3; % then display results with that k. But have not set k-means to replicate (speed concern), can be very noisy
+uicontrol('Parent',tab{i_tab},'Style','text','String','Find best k in range:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight])
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_kmeans_elbow_Callback);
+
+%% UI row 2: Auto-clustering
+i_row = 2;
+i = 1;n = 0;
+
+i=i+n; % Adjacent clusters (arranged by hier.) will be merged
+n=4;  % if correlation between centroids is above merging threshold
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Merge thres. (corr. based)',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_merge_Callback});
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit','String',num2str(thres_merge),...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@edit_mergethres_Callback});
+
+i=i+n; % further split initial clusters so that the average within-cluster corr coeff is above thres
+n=2; % (not so iterative anymore, but could easily restore)
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Iter. split',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_iter_split});
+
+i=i+n;
+n=1; 
+uicontrol('Parent',tab{i_tab},'Style','edit','String',num2str(thres_split),...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@edit_splitthres_Callback});
+
+i=i+n;
+n=3; % minimal size of cluster, otherwise delete at the end
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Cluster size thres.',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_thressize_Callback});
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit','String',num2str(thres_size),...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@edit_sizethres_Callback});
+
+i=i+n+1; % longest script here. Splits clusters and prunes them, to yield only very tight clusters.
+n=3; % really pretty results, but takes a while when regressing with every centroid. Read code for details.
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Full Auto-Clustering',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_autoclus_Callback});
+
+i=i+n;
+n=4; % by default it starts with a k-mean of 20 of the current cells. Could skip that if already clustered.
+uicontrol('Parent',tab{i_tab},'Style','checkbox','String','(starting with k-mean)',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],'Value',1,...
+    'Callback',@checkbox_wkmeans_Callback);
+
+%% UI row 3: misc plots
+i_row = 3;
+i = 1;n = 0;
+
+i=i+n; % k-means are always ranked like in hierachical clustering ~ optimal leaf order
+n=2; % here you can rank them again and plot the dengrogram.
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Hier. plot',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_hierplot_Callback);
+
+i=i+n; % partitioning based on hier. clustering
+n=2; % choose between max cluster numbers...
+uicontrol('Parent',tab{i_tab},'Style','text','String','Hier.cut, max n:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_hierpartn_Callback);
+
+i=i+n; % partitioning based on hier. clustering
+n=3; % ...or set correlation value threshold
+uicontrol('Parent',tab{i_tab},'Style','text','String','Hier.cut, corr thres:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_hierpartthres_Callback);
+
+i=i+n; % hier. partition in place, i.e. without rearranging order clusters
+n=3; 
+uicontrol('Parent',tab{i_tab},'Style','checkbox','String','Hier.cut in place',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],'Value',1,...
+    'Callback',@checkbox_hierinplace_Callback);
+
+i=i+n;
+n=2; % Plots the correlation between all current clusters as a matrix
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Corr. plot',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_corrplot_Callback});
+
+i=i+n;
+n=2; % random plots not really worth consolidating, only for 1-click convenience
+uicontrol('Parent',tab{i_tab},'Style','text','String','temp. plots:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight])
+
+i=i+n;
+n=3; % e.g. this one looks at the history for the fish presented with 4 B/W stimuli
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','plot 4x4(stim)',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_plot4x4_Callback});
+
+%% UI ----- tab five ----- (Saved Clusters)
+i_tab = 5;
+
+%% UI row 1: Class
+i_row = 1; % There is not really a difference between Class and Cluster
+i = 1;n = 0; % I just like to have 2 layers, so I save bigger sets as Class, smaller things as Cluster
+% also now Clusters come in cluster-groups ('Clusgroup'), i.e. multiple groups of Clusters...
+% the saving is sort of a pain though, several Update___ internal functions dedicated to this.
+
+i=i+n;
+n=1; % 
+uicontrol('Parent',tab{i_tab},'Style','text','String','Class:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=4;
+temp = MakeMenu({Class.name});
+hclassmenu = uicontrol('Parent',tab{i_tab},'Style','popupmenu','String',temp,...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@popup_classmenu_Callback);
+
+i=i+n;
+n=1; % just edit and press enter to edit current name
+uicontrol('Parent',tab{i_tab},'Style','text','String','Edit:',... 
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=3;
+hclassname = uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@edit_editclassname_Callback});
+set(hclassname,'String',Class(classID).name);
+
+i=i+n;
+n=2; % save the current cells, with current cluster divisions, as 1 class
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Save class',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_saveclass_Callback);
+
+i=i+n;
+n=1; % not really useful, just to help me organize the Class name
+uicontrol('Parent',tab{i_tab},'Style','text','String','Header:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1; 
+uicontrol('Parent',tab{i_tab},'Style','edit','String','Test',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_class_header_Callback);
+
+i=i+n;
+n=1; % write down new name for new cluster
+uicontrol('Parent',tab{i_tab},'Style','text','String','New:',... 
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=2;
+uicontrol('Parent',tab{i_tab},'Style','edit','String','(blank)',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_newclassname_Callback);
+
+i=i+n;
+n=2; % new class is made when pressing this button
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Make class',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_makeclass_Callback);
+
+i=i+n;
+n=2; % will prompt you to confirm
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Delete Class!',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_delclass_Callback});
+
+%% UI row 2: Cluster
+i_row = 2; % Virtually the same as Class, except added a ranking option
+i = 1;n = 0;
+
+i=i+n;
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','text','String','Cluster (group):',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1;
+num = length(VAR(i_fish).ClusGroup);
+temp = {}; for j = 1:num, temp = [temp,{num2str(j)}];end
+hclusgroupmenu = uicontrol('Parent',tab{i_tab},'Style','popupmenu','String',temp,...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@popup_clusgroupmenu_Callback);
+
+i=i+n;
+n=3;
+temp = MakeMenu({Cluster.name});
+hclusmenu = uicontrol('Parent',tab{i_tab},'Style','popupmenu','String',temp,...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@popup_clusmenu_Callback);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','text','String','Edit:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=3;
+hclusname = uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@edit_editclusname_Callback});
+set(hclusname,'String',Cluster(clusID).name);
+
+i=i+n;
+n=2;
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Save cluster',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_saveclus_Callback);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','text','String','Header:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit','String','Test',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_clus_header_Callback);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','text','String','New:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=2;
+uicontrol('Parent',tab{i_tab},'Style','edit','String','(blank)',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_newclusname_Callback);
+
+i=i+n;
+n=2;
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Make cluster',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@pushbutton_makeclus_Callback);
+
+i=i+n;
+n=2;
+uicontrol('Parent',tab{i_tab},'Style','text','String','Set rank:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@edit_setrank_Callback});
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','text','String','Notes:',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=2;
+uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@edit_notes_Callback});
+
+i=i+n;
+n=2;
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','Delete Cluster!',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_delclus_Callback});
+
+%% UI row 3: misc
+i_row = 3; % continuation, dealing with the Cluster groups 
+i = 1;n = 0; % (Cluster-group number: number menu before the Cluster-name menu)
+
+i=i+n;
+n=2; % Combines the chosen clusters into one view (can save as 1 class or cluster then)
+uicontrol('Parent',tab{i_tab},'Style','text','String','Union(cluster):',... % (eg 1,3-5)
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight]);
+
+i=i+n;
+n=1;
+uicontrol('Parent',tab{i_tab},'Style','edit',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',@edit_clusUnion_Callback);
+
+i=i+n; % just adds a new number to the Clustergroup-number menu, 
+n=3; % and saves current view as the first cluster in the new Clustergroup
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','new Clustergroup',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_newclusgroup_Callback});
+
+i=i+n;
+n=3; % delete current
+uicontrol('Parent',tab{i_tab},'Style','pushbutton','String','del Clustergroup',...
+    'Position',[grid(i) yrow(i_row) bwidth*n rheight],...
+    'Callback',{@pushbutton_delclusgroup_Callback});
+
+%% Load figure
+
+UpdateClusID(hfig,clusID);
+
+%% get local function handles
+
+fcns = localfunctions;
+
+end
+
+%% Callback functions for UI elements: 
+
+%% ----- tab one ----- (General)
+
+%% row 1: File
+
+function pushbutton_save_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+i_fish = getappdata(hfig,'i_fish');
+global VAR;
+VAR(i_fish).Class = getappdata(hfig,'Class');
+VAR(i_fish).ClusGroup = CurrentClusGroup(hfig);
+disp('saved to workspace ''VAR''');
+end
+
+function pushbutton_savemat_Callback(hObject,~)
+disp('saving...');
+hfig = getParentFigure(hObject);
+i_fish = getappdata(hfig,'i_fish');
+arcmatfolder = getappdata(hfig,'arcmatfolder');
+global VAR;
+VAR(i_fish).Class = getappdata(hfig,'Class');
+VAR(i_fish).ClusGroup = CurrentClusGroup(hfig);
+timestamp  = datestr(now,'mmddyy_HHMM');
+matname = [timestamp '.mat'];
+save(fullfile(arcmatfolder,matname),'VAR','-v6');
+save('VAR_current.mat','VAR','-v6');
+disp('saved both to workspace and .mat');
+end
+
+function pushbutton_writeZstack_Callback(hObject,~) 
+disp('preparing z-stack...');
+isMarkAllCells = 1;
+
+hfig = getParentFigure(hObject);
+isfullfish = getappdata(hfig,'isfullfish');
+if ~isfullfish,
+    errordlg('Load full fish first!');
+    return;
+end
+
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+CInfo = getappdata(hfig,'CInfo');
+ave_stack = getappdata(hfig,'ave_stack');
+
+timestamp = datestr(now,'mmddyy_HHMMSS');
+tiffName = ['stack_' timestamp '.tif'];
+
+U = unique(gIX);
+numK = length(U);
+
+% left half: stack with cells marked; 
+% right half: original anatomy, or mark all cells
+
+ave_stack2=zeros(size(ave_stack,1), size(ave_stack,2)*2, size(ave_stack,3) ,3);
+nPlanes=size(ave_stack,3);
+dimv_yxz=size(ave_stack);
+stacklen=numel(ave_stack);
+
+circle=makeDisk2(10,21);
+[r, v]=find(circle);
+r=r-11;v=v-11;
+circle_inds  = r*dimv_yxz(1)+v;
+cmap = hsv(numK);
+weight = 0.3;
+
+for i=1:nPlanes,
+    ave_stack2(:,:,i,:)=repmat(imNormalize99(ave_stack(:,:,i)),[1 2 1 3]);
+end
+
+for j=1:length(cIX)
+    cinds=(CInfo(cIX(j)).center(2)-1)*dimv_yxz(1)+CInfo(cIX(j)).center(1);
+    labelinds=find((cinds+circle_inds)>0 & (cinds+circle_inds)<=dimv_yxz(1)*dimv_yxz(2));
+    zinds=dimv_yxz(1)*dimv_yxz(2)*2*(CInfo(cIX(j)).slice-1);
+    ix = find(U==gIX(j));
+    ixs = cinds+circle_inds(labelinds)+zinds;
+    ave_stack2(ixs)=cmap(ix,1)*weight + ave_stack2(ixs)*(1-weight);
+    ixs = cinds+circle_inds(labelinds)+zinds+stacklen*2;
+    ave_stack2(ixs)=cmap(ix,2)*weight + ave_stack2(ixs)*(1-weight);
+    ixs = cinds+circle_inds(labelinds)+zinds+stacklen*4;
+    ave_stack2(ixs)=cmap(ix,3)*weight + ave_stack2(ixs)*(1-weight);
+end
+if isMarkAllCells,
+    clr = [0,1,0];
+    numcell = getappdata(hfig,'numcell');
+    for j=1:numcell,
+        shift = dimv_yxz(1)*dimv_yxz(2);
+        cinds=(CInfo(j).center(2)-1)*dimv_yxz(1)+CInfo(j).center(1);
+        labelinds=find((cinds+circle_inds)>0 & (cinds+circle_inds)<=dimv_yxz(1)*dimv_yxz(2));
+        zinds=dimv_yxz(1)*dimv_yxz(2)*2*(CInfo(j).slice-1);
+        ixs = cinds+circle_inds(labelinds)+zinds + shift;
+        ave_stack2(ixs)=clr(1)*weight + ave_stack2(ixs)*(1-weight);
+        ixs = cinds+circle_inds(labelinds)+zinds+stacklen*2 + shift;
+        ave_stack2(ixs)=clr(2)*weight + ave_stack2(ixs)*(1-weight);
+        ixs = cinds+circle_inds(labelinds)+zinds+stacklen*4 + shift;
+        ave_stack2(ixs)=clr(3)*weight + ave_stack2(ixs)*(1-weight);
+    end
+end
+h = figure;
+for i_plane = 1:nPlanes,
+    im = squeeze(ave_stack2(:,:,i_plane,:));
+    image(im);
+    % save tiff
+    if (i_plane == 1)
+        imwrite(im, tiffName, 'compression','none','writemode','overwrite')
+    else
+        imwrite(im, tiffName, 'compression','none','writemode','append')
+    end
+    pause(0.2)
+end
+close(h)
+end
+
+function out = makeDisk2(radius, dim)
+center=floor(dim/2)+1;
+out=zeros(dim);
+for x=1:dim
+    for y=1:dim
+        if norm([x,y]-[center,center])<=radius
+            out(x,y)=1;
+        end
+    end
+end
+end
+
+function pushbutton_popupplot_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+
+% same as function RefreshFigure(hfig), except new axes not global
+
+% figure('Position',[50,100,853,512]); % right plot size: 2048x1706
+% h1 = axes('Position',[0, 0, 0.5, 1]); % left ~subplot
+% h2 = axes('Position',[0.50, 0, 0.5, 1]); % right ~subplot
+figure('Position',[50,100,1600,900]);
+h1 = axes('Position',[0.05, 0.03, 0.53, 0.94]); % left ~subplot
+h2 = axes('Position',[0.61, 0.03, 0.35, 0.94]); % right ~subplot
+
+CInfo = getappdata(hfig,'CInfo');
+M = getappdata(hfig,'M');
+dataFR = getappdata(hfig,'dataFR');
+fictive = getappdata(hfig,'fictive');
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+
+numK = getappdata(hfig,'numK');
+stim = getappdata(hfig,'stim');
+anat_yx = getappdata(hfig,'anat_yx');
+anat_yz = getappdata(hfig,'anat_yz');
+anat_zx = getappdata(hfig,'anat_zx');
+isCentroid = getappdata(hfig,'isCentroid');
+clrmap = getappdata(hfig,'clrmap');
+rankscore = getappdata(hfig,'rankscore');
+rankID = getappdata(hfig,'rankID');
+
+iswrite = (rankID>=2);
+
+% left subplot
+axes(h1);
+if isCentroid,
+    [C,~] = FindCentroid(gIX,M);
+    DrawClusters(h1,C,unique(gIX),dataFR,numK,stim,fictive,clrmap,rankscore,iswrite);
+else
+    DrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,iswrite);
+end
+
+% right subplot
+axes(h2);
+DrawClustersOnMap_LSh(CInfo,cIX,gIX,numK,anat_yx,anat_yz,anat_zx,clrmap);
+
+end
+
+function pushbutton_exporttoworkspace_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+assignin('base', 'M', getappdata(hfig,'M'));
+assignin('base', 'cIX', getappdata(hfig,'cIX'));
+assignin('base', 'gIX', getappdata(hfig,'gIX'));
+assignin('base', 'numK', getappdata(hfig,'numK'));
+assignin('base', 'stim', getappdata(hfig,'stim'));
+assignin('base', 'fictive', getappdata(hfig,'fictive'));
+assignin('base', 'M_0_fluo', getappdata(hfig,'M_0_fluo'));
+end
+
+%% row 2: Load
+
+function popup_fishmenu_Callback(hObject,~)
+new_i_fish = get(hObject,'Value');
+hfig = getParentFigure(hObject);
+QuickUpdateFish(hfig,new_i_fish);
+end
+
+function QuickUpdateFish(hfig,new_i_fish,init) %#ok<INUSD>
+CONSTs = getappdata(hfig,'CONSTs');
+setappdata(hfig,'isfullfish',0);
+
+% load all fields from CONSTs, with names preserved
+names = fieldnames(CONSTs{new_i_fish}); % cell of strings
+for i = 1:length(names),
+    setappdata(hfig,names{i},eval(['CONSTs{new_i_fish}.',names{i}]));
+end
+
+% recontruct the 3 big matrices from CIX to original size (numcell)
+CIX = CONSTs{new_i_fish}.CIX;
+numcell = CONSTs{new_i_fish}.numcell;
+
+temp = CONSTs{new_i_fish}.CRAZ;
+CRAZ = zeros(numcell,size(temp,2));
+CRAZ(CIX,:) = temp;
+
+temp = CONSTs{new_i_fish}.CRZt;
+CRZt = zeros(numcell,size(temp,2));
+CRZt(CIX,:) = temp;
+
+temp = CONSTs{new_i_fish}.CInfo;
+CInfo(numcell).center = '';
+CInfo(numcell).area = '';
+CInfo(numcell).slice = '';
+CInfo(numcell).inds = '';
+CInfo(numcell).y_minmax = '';
+CInfo(numcell).x_minmax = '';
+CInfo(CIX) = temp;
+
+setappdata(hfig,'CRAZ',CRAZ); % Matrix array, misc collection
+setappdata(hfig,'CRZt',CRZt); % Matrix array, misc collection
+setappdata(hfig,'CInfo',CInfo); % Matrix array, misc collection
+
+UpdateFishData(hfig,new_i_fish);
+if ~exist('init','var'),
+    UpdateFishDisplay(hfig,new_i_fish);
+end
+end
+
+function UpdateFishData(hfig,new_i_fish)
+UpdateDatamenu_Direct(hfig,1);
+
+% save ClusGroup before updating, if applicable
+clusgroupID = getappdata(hfig,'clusgroupID');
+if ~isempty(clusgroupID),
+    new_clusgroupID = 1;
+    UpdateClusGroupID(hfig,clusgroupID,new_clusgroupID,'norefresh'); % to save ClusGroup
+end
+% set new i_fish after UpdateClusGroupID, which saves into old 'i_fish'
+setappdata(hfig,'i_fish',new_i_fish);
+
+% load from VAR
+global VAR;
+Class = VAR(new_i_fish).Class; % actually collection of classes, each class can have many groups
+setappdata(hfig,'Class',Class);
+
+ClusGroup = VAR(new_i_fish).ClusGroup; % group of 'Cluster', i.e. 2nd level collections
+setappdata(hfig,'ClusGroup',ClusGroup);
+
+clusgroupID = 1;
+setappdata(hfig,'clusgroupID',clusgroupID);
+
+Cluster = ClusGroup{clusgroupID};% collection of clusters, 'Cluster' structurally same as 'Class'
+setappdata(hfig,'Cluster',Cluster);
+end
+
+function UpdateFishDisplay(hfig,new_i_fish)
+stim = getappdata(hfig,'stim');
+M_fish_set = getappdata(hfig,'M_fish_set');
+
+fishset = M_fish_set(new_i_fish);
+[~, names] = GetStimRegressor(stim,fishset);
+global hstimreg;
+set(hstimreg,'String',['(choose)',(names)]);
+
+% update display
+UpdateClassID(hfig,1,'norefresh');
+clusgroupID = 1;
+new_clusgroupID = 1;
+UpdateClusGroupID(hfig,clusgroupID,new_clusgroupID); % to display new menu
+end
+
+function popup_loadfullfishmenu_Callback(hObject,~)
+new_i_fish = get(hObject,'Value')-1;
+if new_i_fish>0,
+    hfig = getParentFigure(hObject);
+    LoadFullFish(hfig,new_i_fish);
+end
+global hfishnum;
+set(hfishnum,'Value',new_i_fish);
+end
+
+function LoadFullFish(hfig,new_i_fish,CONST)
+data_dir=getappdata(hfig,'data_dir');
+if ~exist('CONST','var'),
+    disp(['loading fish #' num2str(new_i_fish) '...']);
+    tic
+    load(fullfile(data_dir,['CONST_F' num2str(new_i_fish) '.mat']),'CONST');
+    toc
+end
+setappdata(hfig,'isfullfish',1);
+
+% load all fields from CONST, with names preserved
+names = fieldnames(CONST); % cell of strings
+for i = 1:length(names),
+    setappdata(hfig,names{i},eval(['CONST.',names{i}]));
+end
+setappdata(hfig,'numcell',length(CONST.CInfo));
+    
+UpdateFishData(hfig,new_i_fish);
+UpdateFishDisplay(hfig,new_i_fish);
+end
+
+function popup_datamenu_Callback(hObject,~)
+ID = get(hObject,'Value');
+hfig = getParentFigure(hObject);
+UpdateDatamenu_Direct(hfig,ID);
+SetDataFR(hfig,1); % also set M (without updating cIX gIX)
+RefreshFigure(hfig);
+end
+
+function UpdateDatamenu_Direct(hfig,ID)
+CRAZ = getappdata(hfig,'CRAZ');
+CRZt = getappdata(hfig,'CRZt');
+FcAvr = getappdata(hfig,'FcAvr');
+Fc = getappdata(hfig,'Fc');
+datanames = getappdata(hfig,'datanames');
+tlists = getappdata(hfig,'tlists');
+photostate = getappdata(hfig,'photostate');
+
+%% SWITCH LEFT/RIGHT DIRECTION AGAIN
+FcAvr = vertcat(FcAvr(2,:),FcAvr(1,:),FcAvr(3:end,:));
+Fc = vertcat(Fc(2,:),Fc(1,:),Fc(3:end,:));
+% now 1=left, 2=right, 3=forward, 4 = raw left, 5 = raw right
+
+%%
+if ID == 1,
+    M_0 = CRAZ;
+    fictive = FcAvr;
+elseif ID == 2,
+    M_0 = CRZt;
+    fictive = Fc;
+else    
+    IX = tlists{ID};
+    M_0 = CRZt(:,IX);
+    temp = Fc(:,IX);
+    % normalize fictive channels, in 2 sets
+    for k = 1:3,%size(F,1),
+        m = temp(k,:);
+        temp(k,:) = (m-min(m))/(max(m)-min(m));
+    end
+    m = temp(4:5,:);
+    temp(4:5,:) = (m-min(min(m)))/(max(max(m))-min(min(m)));
+    fictive = temp;
+end
+% stim from photostate
+stim = photostate(tlists{ID}); 
+
+setappdata(hfig,'M_0_fluo',M_0);
+setappdata(hfig,'fictive',fictive);
+setappdata(hfig,'fictive_0',fictive);
+setappdata(hfig,'dataname',datanames(ID));
+setappdata(hfig,'stim',stim);
+
+global hdatamenu;
+if ~isempty(hdatamenu), % before GUI initialization
+    % 'global' remnant from previous run: hdatamenu is not empty but is invalid object
+    if isvalid(hdatamenu), 
+        set(hdatamenu,'String',datanames);
+        set(hdatamenu,'Value',ID);
+    end
+end
+end
+
+function checkbox_showcentroids_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+setappdata(hfig,'isCentroid',get(hObject,'Value'));
+RefreshFigure(hfig);
+end
+
+%% ----- tab two ----- (Operations)
+
+%% row 1: Range
+
+function pushbutton_back_Callback(hObject,~)
+global hback hfwd;
+hfig = getParentFigure(hObject);
+bC = getappdata(hfig,'bCache');
+fC = getappdata(hfig,'fCache');
+
+if ~isempty(bC.cIX{1}),
+    % set last into forward cache
+    fC.cIX = [getappdata(hfig,'cIX'),fC.cIX];
+    fC.gIX = [getappdata(hfig,'gIX'),fC.gIX];
+    fC.numK = [getappdata(hfig,'numK'),fC.numK];
+    % retrieve
+    cIX = bC.cIX{1};
+    gIX = bC.gIX{1};
+    numK = bC.numK{1};
+    bC.cIX(1) = [];
+    bC.gIX(1) = [];
+    bC.numK(1) = [];
+    % set M
+    M_0 = GetM_0(hfig);
+    M = M_0(cIX,:);
+    % save
+    setappdata(hfig,'M',M);
+    setappdata(hfig,'bCache',bC);
+    setappdata(hfig,'fCache',fC);
+    setappdata(hfig,'cIX',cIX);
+    setappdata(hfig,'gIX',gIX);
+    setappdata(hfig,'numK',numK);
+    % handle rankID: >=2 means write numbers as text next to colorbar
+    setappdata(hfig,'rankID',0);
+    % finish
+    disp('back (from cache)')
+    RefreshFigure(hfig);
+    set(hfwd,'enable','on');
+else % nothing to retrieve
+    set(hback,'enable','off');
+end
+end
+
+function pushbutton_forward_Callback(hObject,~)
+global hback hfwd;
+hfig = getParentFigure(hObject);
+bC = getappdata(hfig,'bCache');
+fC = getappdata(hfig,'fCache');
+
+if ~isempty(fC.cIX{1}),
+    % set last into (backward) cache
+    bC.cIX = [getappdata(hfig,'cIX'),bC.cIX];
+    bC.gIX = [getappdata(hfig,'gIX'),bC.gIX];
+    bC.numK = [getappdata(hfig,'numK'),bC.numK];
+    % retrieve
+    cIX = fC.cIX{1};
+    gIX = fC.gIX{1};
+    numK = fC.numK{1};
+    fC.cIX(1) = [];
+    fC.gIX(1) = [];
+    fC.numK(1) = [];
+    % set M
+    M_0 = GetM_0(hfig);
+    M = M_0(cIX,:);
+    % save
+    setappdata(hfig,'M',M);
+    setappdata(hfig,'bCache',bC);
+    setappdata(hfig,'fCache',fC);
+    setappdata(hfig,'cIX',cIX);
+    setappdata(hfig,'gIX',gIX);
+    setappdata(hfig,'numK',numK);
+    % handle rankID: >=2 means write numbers as text next to colorbar
+    setappdata(hfig,'rankID',0);
+    % finish
+    disp('forward (from cache)')
+    RefreshFigure(hfig);
+    set(hback,'enable','on');
+else
+    set(hfwd,'enable','off');
+end
+end
+
+function edit_choose_range_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+% get/format range
+str = get(hObject,'String'); 
+if ~isempty(str),
+    str = strrep(str,'end',num2str(max(gIX)));
+    range = ParseRange(str);
+    % update indices
+    tempI = [];
+    for i = range,
+        tempI = [tempI;find(gIX==i)];
+    end
+    cIX = cIX(tempI);
+    gIX = gIX(tempI);
+    UpdateIndices(hfig,cIX,gIX);
+    RefreshFigure(hfig);
+end
+end
+
+function range = ParseRange(str)
+str = strrep(str,':','-'); % e.g. str= '1,3,5:8';
+C = textscan(str,'%d','delimiter',',');
+m = C{:};
+range = [];
+for i = 1:length(m),
+    if m(i)>0,
+        range = [range,m(i)]; %#ok<AGROW>
+    else % have '-'sign,
+        range = [range,m(i-1)+1:-m(i)]; %#ok<AGROW>
+    end
+end
+end
+
+function edit_exclude_range_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+% get/format range
+str = get(hObject,'String');
+if ~isempty(str),
+    str = strrep(str,'end',num2str(max(gIX)));
+    range = ParseRange(str);
+    range = setdiff(unique(gIX),range)';
+    % update indices
+    tempI = [];
+    for i = range,
+        tempI = [tempI;find(gIX==i)];
+    end
+    cIX = cIX(tempI);
+    gIX = gIX(tempI);
+    UpdateIndices(hfig,cIX,gIX);
+    RefreshFigure(hfig);
+end
+end
+
+function edit_fuse_range_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+% get/format range
+str = get(hObject,'String');
+if ~isempty(str),
+    str = strrep(str,'end',num2str(max(gIX)));
+    range = ParseRange(str);
+    % update indices
+    tempI = [];
+    for i = range,
+        tempI = [tempI;find(gIX==i)];
+    end
+    gIX(tempI) = gIX(tempI(1));
+    [gIX, numK] = SqueezeGroupIX(gIX);
+    UpdateIndices(hfig,cIX,gIX,numK);
+    RefreshFigure(hfig);
+end
+end
+
+function edit_t_range_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+M = getappdata(hfig,'M');
+
+% get/format range
+str = get(hObject,'String'); 
+if ~isempty(str),
+    str = strrep(str,'end',num2str(size(M,2)));
+    range = ParseRange(str);
+    
+    UpdateTRange(hfig,range)
+    RefreshFigure(hfig);
+end
+end
+
+function UpdateTRange(hfig,range)
+M_0 = GetM_0(hfig);
+cIX = getappdata(hfig,'cIX');
+fictive_0 = getappdata(hfig,'fictive_0');
+% photostate_0 = getappdata(hfig,'photostate_0');
+
+% set M
+M = M_0(cIX,range);
+setappdata(hfig,'M',M);
+% set fictive
+fictive = fictive_0(:,range);
+setappdata(hfig,'fictive',fictive);
+%     % set photostate
+%     photostate = photostate_0(:,range);
+%     setappdata(hfig,'photostate',photostate);
+
+end
+
+%% row 2: Operations
+
+function popup_operations_Callback(hObject,~)
+opID = get(hObject,'Value') - 1;
+hfig = getParentFigure(hObject);
+setappdata(hfig,'opID',opID);
+% highlight UI
+global hopID;
+set(hopID,'BackgroundColor',[1,1,0.8]);
+end
+
+function popup_ranking_Callback(hObject,~)
+% menu = {'(ranking)','hier.','size','stim-lock','corr','noise'};
+rankID = get(hObject,'Value') - 1;
+hfig = getParentFigure(hObject);
+setappdata(hfig,'rankID',rankID);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+M = getappdata(hfig,'M');
+
+[gIX, numU] = SqueezeGroupIX(gIX);
+switch rankID,
+    case 1,
+        disp('hier. (default)');
+        [gIX, numU] = HierClus(M,gIX);
+    case 2,
+        disp('size');
+        H = zeros(numU,1);
+        for i = 1:numU,
+            H(i) = length(find(gIX==i));
+        end
+        [gIX,rankscore] = SortH(H,gIX,numU,'descend');
+    case 3,
+        disp('stim-lock');
+        [gIX,rankscore] = RankByStimLock_Direct(hfig,cIX,gIX,M,numU);
+    case 4,
+        disp('corr');
+        [~,D] = FindCentroid(gIX,M);
+        [gIX,rankscore] = SortH(D,gIX,numU);
+        rankscore = 1-rankscore;
+    case 5,
+        disp('motor');
+        [gIX,rankscore] = RankByMotorStim_Direct(hfig,gIX,M,numU,1);
+    case 6,
+        disp('motor');
+        [gIX,rankscore] = RankByMotorStim_Direct(hfig,gIX,M,numU,2);
+    case 7,
+        disp('motor');
+        [gIX,rankscore] = RankByMotorStim_Direct(hfig,gIX,M,numU,3);
+    case 8,
+        disp('motor');
+        [gIX,rankscore] = RankByMotorStim_Direct(hfig,gIX,M,numU,4);
+    case 9,
+        disp('stim-motor');
+        [~,rankscore1] = RankByStimLock_Direct(hfig,cIX,gIX,M,numU);
+        [~,rankscore2] = RankByMotorStim_Direct(hfig,gIX,M,numU,1);
+        rankscore = rankscore1 - rankscore2;
+end
+if rankID>1,
+    setappdata(hfig,'rankscore',round(rankscore*100)/100);
+    setappdata(hfig,'clrmap','jet');
+else
+    setappdata(hfig,'clrmap','hsv');
+end
+UpdateIndices(hfig,cIX,gIX,numU);
+RefreshFigure(hfig);
+disp('ranking complete');
+end
+
+function [gIX,rankscore] = RankByStimLock_Direct(hfig,cIX,gIX,M,numU)
+periods = getappdata(hfig,'periods');
+
+if length(periods)==1,
+    period = periods{1};
+    [C,~] = FindCentroid(gIX,M);
+else % i_fish>=8,
+    tlists = getappdata(hfig,'tlists');
+    CRZt = getappdata(hfig,'CRZt');
+    IX = tlists{6}; % ptomr_circ
+    M_ = CRZt(cIX,IX);
+    [C,~] = FindCentroid(gIX,M_);
+    periods = getappdata(hfig,'periods');
+    period = periods{1}+periods{2};
+end
+% C2 = zscore(C,0,2);
+% C_3D = reshape(C2,size(C2,1),period,[]);
+C_3D_0 = reshape(C,size(C,1),period,[]);
+C_3D = zscore(C_3D_0,0,2);
+
+H = nanmean(nanstd(C_3D,0,3),2);
+[gIX,rankscore] = SortH(H,gIX,numU);
+end
+
+function [gIX,rankscore] = RankByMotorStim_Direct(hfig,gIX,M,numU,option)
+C = FindCentroid(gIX,M);
+fictive = getappdata(hfig,'fictive');
+
+reg = zeros(3,length(fictive));
+reg(1,:) = fictive(5,:); % Left
+reg(2,:) = fictive(4,:); % Right
+reg(3,:) = mean(fictive(4:5,:));
+H = zeros(numU,1);
+shift = zeros(numU,1);
+a = zeros(1,3);
+I = zeros(1,3);
+for i = 1:numU,
+    switch option,
+        case 1,
+            for j = 1:3,
+                [a(j),I(j)] = max(abs(xcorr(C(i,:),reg(j,:),'coeff')));
+            end
+            [H(i),jmax] = max(a);
+            shift(i) = I(jmax) - length(fictive);
+        case 2,
+            [H(i),I] = max(xcorr(C(i,:),reg(1,:),'coeff'));
+            shift(i) = I - length(fictive);
+        case 3,
+            [H(i),I] = max(xcorr(C(i,:),reg(2,:),'coeff'));
+            shift(i) = I - length(fictive);
+        case 4,
+            [H(i),I] = max(xcorr(C(i,:),reg(3,:),'coeff'));
+            shift(i) = I - length(fictive);
+    end
+end
+[gIX,rankscore] = SortH(H,gIX,numU,'descend');
+assignin('base', 'shift', shift);
+end
+
+function [gIX,B] = SortH(H,gIX,numU,descend) % new gIX is sorted based on H, size(H)=[numU,1];
+if exist('descend','var'),
+    [B,I] = sort(H,'descend');
+else
+    [B,I] = sort(H);
+end
+gIX_last = gIX;
+for i = 1:numU,
+    gIX(gIX_last==I(i)) = i;
+end
+end
+
+function pushbutton_sqeeze_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+[gIX, numU] = SqueezeGroupIX(gIX);
+UpdateIndices(hfig,cIX,gIX,numU);
+RefreshFigure(hfig);
+end
+
+function pushbutton_flipud_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+
+U = unique(gIX);
+U_ = flipud(U);
+
+gIX_ = gIX;
+for i = 1:length(U),
+    gIX_(gIX==U(i)) = U_(i);
+end
+
+UpdateIndices(hfig,cIX,gIX_);
+RefreshFigure(hfig);
+end
+
+function pushbutton_clrmap_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+clrmap = getappdata(hfig,'clrmap');
+if strcmp(clrmap,'jet'),
+    setappdata(hfig,'clrmap','hsv');
+else
+    setappdata(hfig,'clrmap','jet');
+end
+RefreshFigure(hfig);
+end
+
+%% row 3: Anatomy
+
+function pushbutton_polygon_yx_Callback(hObject,~)
+k_zres = 20;
+
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+% M = getappdata(hfig,'M');
+CInfo = getappdata(hfig,'CInfo');
+numcell = getappdata(hfig,'numcell');
+anat_yx = getappdata(hfig,'anat_yx');
+anat_zx = getappdata(hfig,'anat_zx');
+dimv_yx = size(anat_yx);
+dimv_zx = size(anat_zx);
+isfullfish = getappdata(hfig,'isfullfish');
+
+h_poly_yx = impoly;
+wait(h_poly_yx); % double click to finalize position!
+% update finalized polygon in bright color
+setColor(h_poly_yx,[0 1 1]);
+
+if isfullfish,
+    IJs = reshape([CInfo.center],2,[])';
+else % Matlab can't handle CInfo with all the empty entries -> manual padding
+    IJs_ = reshape([CInfo(cIX).center],2,[])';
+    IJs = ones(numcell,2);
+    IJs(cIX,:) = IJs_;
+end
+A = sub2ind(dimv_yx(1:2),IJs(:,1),IJs(:,2));
+MaskArray = createMask(h_poly_yx);
+MaskArray(1:dimv_zx*k_zres+10,:) = [];
+B = find(MaskArray); % find indices of pixels within ROI
+
+temp = find(ismember(A,B));
+[IX,ia,~] = intersect(cIX,temp);
+
+cIX = IX;
+gIX = gIX(ia);
+
+UpdateIndices(hfig,cIX,gIX);
+RefreshFigure(hfig);
+end
+
+function pushbutton_polygon_yz_Callback(hObject,~)
+k_zres = 20;
+
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+% M = getappdata(hfig,'M');
+CInfo = getappdata(hfig,'CInfo');
+numcell = getappdata(hfig,'numcell');
+anat_yx = getappdata(hfig,'anat_yx');
+anat_yz = getappdata(hfig,'anat_yz');
+anat_zx = getappdata(hfig,'anat_zx');
+dimv_yx = size(anat_yx);
+dimv_yz = size(anat_yz);
+dimv_zx = size(anat_zx);
+isfullfish = getappdata(hfig,'isfullfish');
+
+h_poly_z = impoly;
+wait(h_poly_z); % double click to finalize position!
+% update finalized polygon in bright color
+setColor(h_poly_z,[0 1 1]);
+
+if isfullfish,
+        IJs = reshape([CInfo.center],2,[])';
+        Zs = [CInfo.slice]';
+else % Matlab can't handle CInfo with all the empty entries -> manual padding
+    IJs_ = reshape([CInfo(cIX).center],2,[])';
+    IJs = ones(numcell,2);
+    IJs(cIX,:) = IJs_;
+    Zs_ = [CInfo(cIX).slice]';
+    Zs = ones(numcell,1);
+    Zs(cIX) = Zs_;
+end
+A = sub2ind(dimv_yz(1:2),IJs(:,1),Zs);
+
+MaskArray = createMask(h_poly_z);
+MaskArray(1:dimv_zx*k_zres+10,:) = [];
+MaskArray(:,1:dimv_yx(2)+10) = [];
+zMaskArray = imresize(MaskArray,[dimv_yz(1),dimv_yz(2)]);
+B = find(zMaskArray); % find indices of pixels within ROI
+
+temp = find(ismember(A,B));
+[IX,ia,~] = intersect(cIX,temp);
+
+cIX = IX;
+gIX = gIX(ia);
+
+UpdateIndices(hfig,cIX,gIX);
+RefreshFigure(hfig);
+end
+
+function pushbutton_polygon_zx_Callback(hObject,~)
+k_zres = 20;
+
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+% M = getappdata(hfig,'M');
+CInfo = getappdata(hfig,'CInfo');
+numcell = getappdata(hfig,'numcell');
+% anat_yx = getappdata(hfig,'anat_yx');
+anat_zx = getappdata(hfig,'anat_zx');
+% dimv_yx = size(anat_yx);
+dimv_zx = size(anat_zx);
+isfullfish = getappdata(hfig,'isfullfish');
+
+h_poly_z = impoly;
+wait(h_poly_z); % double click to finalize position!
+% update finalized polygon in bright color
+setColor(h_poly_z,[0 1 1]);
+
+if isfullfish,
+        IJs = reshape([CInfo.center],2,[])';
+        Zs = [CInfo.slice]';
+else % Matlab can't handle CInfo with all the empty entries -> manual padding
+    IJs_ = reshape([CInfo(cIX).center],2,[])';
+    IJs = ones(numcell,2);
+    IJs(cIX,:) = IJs_;
+    Zs_ = [CInfo(cIX).slice]';
+    Zs = ones(numcell,1);
+    Zs(cIX) = Zs_;
+end
+A = sub2ind(dimv_zx(1:2),Zs,IJs(:,2));
+
+MaskArray = createMask(h_poly_z);
+
+MaskArray(dimv_zx(1)*k_zres+1:end,:) = [];
+MaskArray(:,dimv_zx(2)+1:end) = [];
+zMaskArray = imresize(MaskArray,[dimv_zx(1),dimv_zx(2)]);
+zMaskArray = flipud(zMaskArray); % careful!
+B = find(zMaskArray); % find indices of pixels within ROI
+
+temp = find(ismember(A,B));
+[IX,ia,~] = intersect(cIX,temp);
+
+cIX = IX;
+gIX = gIX(ia);
+
+UpdateIndices(hfig,cIX,gIX);
+RefreshFigure(hfig);
+end
+
+function pushbutton_withinConvexHull_Callback(hObject,~) % only works with full load
+disp('find points within boundary...');
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+CInfo = getappdata(hfig,'CInfo');
+IJs = reshape([CInfo.center],2,[])';
+Zs = [CInfo.slice]';
+% point sets
+xyz_all = horzcat(IJs, Zs);
+xyz_cIX = horzcat(IJs(cIX,:), Zs(cIX));
+
+tri = delaunayn(xyz_cIX); % Generate delaunay triangulization
+t = tsearchn(xyz_cIX, tri, xyz_all); % Determine which triangle point is within
+I_inside = ~isnan(t);
+
+cIX_1 = cIX;
+gIX_1 = gIX;
+cIX_2 = setdiff(find(I_inside),cIX_1);
+gIX_2 = ones(length(cIX_2),1)*(max(gIX_1)+1);
+cIX = [cIX_1;cIX_2];
+gIX = [gIX_1;gIX_2];
+
+UpdateIndices(hfig,cIX,gIX,length(unique(gIX)));
+RefreshFigure(hfig);
+disp('search complete');
+end
+
+%% ----- tab three ----- (Regression)
+
+%% row 1: regressor
+
+function popup_getstimreg_Callback(hObject,~)
+i_reg = get(hObject,'Value')-1;
+hfig = getParentFigure(hObject);
+if i_reg==0,
+    return;
+end
+setappdata(hfig,'regchoice',{1,i_reg});
+% highlight the choice (yellow)
+global hstimreg hmotorreg hcentroidreg;
+set(hstimreg,'BackgroundColor',[1,1,0.8]); % yellow
+set(hmotorreg,'BackgroundColor',[1,1,1]);
+set(hcentroidreg,'BackgroundColor',[1,1,1]);
+
+isPlotReg = getappdata(hfig,'isPlotReg');
+if isPlotReg,
+    PlotRegWithStimMotor(hfig);
+end
+end
+
+function PlotRegWithStimMotor(hfig)
+stim = getappdata(hfig,'stim');
+fictive = getappdata(hfig,'fictive');
+regressor = GetRegressor(hfig);
+% plot regressor
+figure;
+halfbarheight = 1;
+stimbar = GetStimBar(halfbarheight,stim);
+reg = imNormalize99(regressor);
+regbar = repmat(reg,[1,1,3]);
+
+subplot(3,1,1);image(stimbar); axis off; title('stimulus');
+subplot(3,1,2);image(regbar); axis off; title('regressor');
+subplot(3,1,3);imagesc(fictive);colormap hot; axis off; title('motor');
+end
+
+function out=imNormalize99(im)
+im=double(im);
+temp=sort(im(:),'descend');
+th1=temp(round(length(im(:))/100));
+th2=min(im(:));
+
+out=(im-th2)/(th1-th2);
+out(out>1)=1;
+end
+
+function popup_getmotorreg_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+i_reg = get(hObject,'Value')-1;
+% {'left swims','right swims','forward swims','raw left','raw right','raw average'};
+if i_reg>0,
+    setappdata(hfig,'regchoice',{2,i_reg});
+end
+% highlight the choice
+global hstimreg hmotorreg hcentroidreg;
+set(hstimreg,'BackgroundColor',[1,1,1]);
+set(hmotorreg,'BackgroundColor',[1,1,0.8]); % yellow
+set(hcentroidreg,'BackgroundColor',[1,1,1]);
+
+isPlotReg = getappdata(hfig,'isPlotReg');
+if isPlotReg,
+    PlotRegWithStimMotor(hfig);
+end
+end
+
+function checkbox_popupplotreg_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+setappdata(hfig,'isPlotReg',get(hObject,'Value'));
+end
+
+function edit_ctrdID_as_reg_Callback(hObject,~)
+str = get(hObject,'String');
+if ~isempty(str),
+    temp = textscan(str,'%d');
+    ctrdID = temp{:};
+end
+hfig = getParentFigure(hObject);
+setappdata(hfig,'regchoice',{3,ctrdID});
+% highlight the choice
+global hstimreg hmotorreg hcentroidreg;
+set(hstimreg,'BackgroundColor',[1,1,1]);
+set(hmotorreg,'BackgroundColor',[1,1,1]);
+set(hcentroidreg,'BackgroundColor',[1,1,0.8]); % yellow
+end
+
+function checkbox_flipstim_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+setappdata(hfig,'isflipstim',get(hObject,'Value'));
+end
+
+%% row 2: regression
+
+function regressor = GetRegressor(hObject)
+hfig = getParentFigure(hObject);
+regchoice = getappdata(hfig,'regchoice');
+stim = getappdata(hfig,'stim');
+    
+if regchoice{1}==1, % stim Regressor
+    i_fish = getappdata(hfig,'i_fish');
+    M_fish_set = getappdata(hfig,'M_fish_set');
+    fishset = M_fish_set(i_fish);
+    [regressors, names] = GetStimRegressor(stim,fishset);    
+    regressor = regressors(regchoice{2}).im;
+    
+elseif regchoice{1}==2, % motor Regressor
+    fictive = getappdata(hfig,'fictive');
+    regressors = GetMotorRegressor(fictive);
+    regressor = regressors(regchoice{2}).im;
+    
+else % regchoice{1}==3, from Centroid
+    ctrdID = regchoice{2};
+    isflipstim = getappdata(hfig,'isflipstim');
+    M = getappdata(hfig,'M');
+    gIX = getappdata(hfig,'gIX');
+    
+    i = find(unique(gIX)==ctrdID);
+    if isempty(i),
+        disp('input is empty!');beep;
+        regressor = [];
+        return;
+    end
+    [C,~] = FindCentroid(gIX,M);
+    regressor = C(i,:);
+    
+    if isflipstim, % swap left/right stim
+        % 0 = all black; 1 = black/white; 2 = white/black; 3 = all white; 4 = all gray;
+        % 10 = forward grating (very slow, more for calibration)
+        % 11 = rightward grating
+        % 12 = leftward grating
+        % swap: 1~2,11~12
+        
+        %     if i_fish == 6 || i_fish == 7,
+        %         halflength = length(stim)/2;
+        %         stim = stim(1:halflength); % length is always even
+        %     end
+        flipstim = stim;
+        flipstim(stim==1)=2;
+        flipstim(stim==2)=1;
+        flipstim(stim==11)=12;
+        flipstim(stim==12)=11;
+        
+        phototrans = GetPhotoTrans(stim);
+        fliptrans = GetPhotoTrans(flipstim);
+        
+        % transition e.g.:
+        % photostate: 2     3     1     3     3     0     0     1     1     0     3     2     0     2     2     1
+        % phototrans: 6    11    13     7    15    12     0     1     5     4     3    14     8     2    10     9
+        
+        IX = [];
+        targetstates = unique(fliptrans,'stable');
+        for i = 1:length(targetstates),
+            IX = [IX, find(phototrans==targetstates(i))];
+        end
+        %     if i_fish == 6 || i_fish == 7,
+        %         reg = regressor(1:halflength);
+        %         reg = reg(IX);
+        %         regressor = repmat(reg,1,2);
+        %     else
+        regressor = regressor(IX);
+        %     end
+    end    
+end
+
+end
+
+function checkbox_plotcorrhist_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+setappdata(hfig,'isPlotCorrHist',get(hObject,'Value'));
+end
+
+function pushbutton_topnum_regression_Callback(hObject,~)
+disp('regression...');
+hfig = getParentFigure(hObject);
+M_0 = getappdata(hfig,'M_0_fluo');
+numTopCorr = getappdata(hfig,'numTopCorr');
+regressor = GetRegressor(hfig);
+
+%% for each cell, find correlation coeff
+M_corr = corr(regressor',M_0');%cell_resp_ave');
+% M_corr_ctrl = corr(regressor.ctrl',M_0');
+
+[~,I] = sort(M_corr,'descend');
+cIX = I(1:numTopCorr)';
+gIX = ceil((1:length(cIX))'/length(cIX)*min(20,length(cIX)));
+
+score = M_corr(cIX);
+rankscore = zeros(1,20);
+for i = 1:20,
+    ix = find(gIX==i);
+    rankscore(i) = mean(score(ix));
+end
+setappdata(hfig,'rankscore',round(rankscore*100)/100);
+setappdata(hfig,'rankID',99);
+
+UpdateIndices(hfig,cIX,gIX,20);
+RefreshFigure(hfig);
+end
+
+function edit_topCorrNumber_Callback(hObject,~)
+str = get(hObject,'String');
+if ~isempty(str),
+    temp = textscan(str,'%d');
+    numTopCorr = temp{:};
+    hfig = getParentFigure(hObject);
+    setappdata(hfig,'numTopCorr',numTopCorr);
+end
+end
+
+function pushbutton_thres_regression_Callback(hObject,~) 
+disp('regression...');
+hfig = getParentFigure(hObject);
+M_0 = getappdata(hfig,'M_0_fluo');
+thres_reg = getappdata(hfig,'thres_reg');
+regressor = GetRegressor(hfig);
+if isempty(regressor),
+    return;
+end
+
+isPlotCorrHist = getappdata(hfig,'isPlotCorrHist');
+
+[cIX,gIX] = Regression_direct(M_0,thres_reg,regressor,isPlotCorrHist);
+
+if ~isempty(gIX),
+    gIX = ceil((1:length(cIX))'/length(cIX)*min(20,length(cIX)));
+    UpdateIndices(hfig,cIX,gIX,40);
+    RefreshFigure(hfig);
+end
+end
+
+function [cIX_,gIX_,R] = Regression_direct(M_0,thres_reg,regressor,isPlotCorrHist) % gIX_ is just ones
+R = corr(regressor',M_0');
+if thres_reg>0,
+    cIX_ = (find(R>thres_reg))';
+elseif thres_reg<0,
+    cIX_ = (find(R<thres_reg))';
+end
+if isempty(cIX_),
+    disp('result is empty!');beep;
+    gIX_ = [];
+    return;
+end
+[~,I] = sort(R(cIX_),'descend');
+cIX_ = cIX_(I);
+gIX_ = ones(size(cIX_));
+%%
+if exist('isPlotCorrHist','var'),
+    if isPlotCorrHist,
+        figure('Position',[500,200,500,150]);
+        hold on;
+        bins = -1:0.025:1;
+        [N,~] = histcounts(R,bins);
+        histogram(R,bins,'FaceColor',[0.4 0.4 0.4]);%,'EdgeColor','none'
+        plot([thres_reg,thres_reg],[0,max(N)],'r--');
+        xlim([-1,1]);ylim([0,max(N)]);
+    end
+end
+end
+
+function edit_regthres_Callback(hObject,~)
+str = get(hObject,'String');
+if ~isempty(str),
+    temp = textscan(str,'%f');
+    thres_reg = temp{:};
+    hfig = getParentFigure(hObject);
+    setappdata(hfig,'thres_reg',thres_reg);
+end
+end
+
+function pushbutton_AllCentroidRegression_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+M_0 = getappdata(hfig,'M_0_fluo');
+thres_reg = getappdata(hfig,'thres_reg');
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+
+[cIX,gIX,numK] = AllCentroidRegression_Direct(M_0,thres_reg,cIX,gIX);
+
+UpdateIndices(hfig,cIX,gIX,numK);
+RefreshFigure(hfig);
+
+disp('all regression complete');
+end
+
+function [cIX,gIX,numK] = AllCentroidRegression_Direct(M_0,thres_reg,cIX,gIX)
+U = unique(gIX);
+clusters(length(U)).cIX = [];
+for i = 1:length(U),
+    disp(['i = ' num2str(i)]);
+    IX = find(gIX == U(i));    
+    [~,regressor] = kmeans(M_0(cIX(IX),:),1,'distance','correlation'); %#ok<FNDSB>
+    cIX_ = Regression_direct(M_0,thres_reg,regressor);
+    clusters(i).cIX = cIX_;
+end
+disp('union...');
+CIX = vertcat(clusters(1:end).cIX);
+GIX = [];
+for i = 1:numel(clusters),
+    GIX = [GIX; ones(length(clusters(i).cIX),1)*i]; %#ok<AGROW>
+end
+[cIX,gIX,numK] = SmartUnique(CIX,GIX,M_0(CIX,:));
+end
+
+function pushbutton_IterCentroidRegression_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+M_0 = getappdata(hfig,'M_0_fluo');
+thres_reg = getappdata(hfig,'thres_reg');
+M = getappdata(hfig,'M');
+gIX = getappdata(hfig,'gIX');
+regchoice = getappdata(hfig,'regchoice');
+if regchoice{1}==3, % from Centroid
+    ctrdID = regchoice{2};
+end
+
+i = find(unique(gIX)==ctrdID);
+if isempty(i),
+    disp('input is empty!');beep;
+    return;
+end
+[C,~] = FindCentroid(gIX,M);
+regressor = C(i,:);
+[cIX_,gIX_] = Regression_direct(M_0,thres_reg,regressor);
+numC = length(cIX_);
+regressor_last = regressor;
+
+if ~isempty(cIX_),
+    for itr = 1:20,
+        [cIX_,gIX_] = Regression_direct(M_0,thres_reg,regressor_last);
+        numC = length(cIX_);
+        disp(num2str(numC));
+        M = M_0(cIX_,:);
+        regressor = FindCentroid(gIX_,M);
+        if itr>1 && abs(numC-numC_last)<5,%corr(regressor_last',regressor')>thres_iter,
+            disp(['converged at itr = ' num2str(itr)]);
+            break;
+        end
+        regressor_last = regressor;
+        numC_last = numC;
+    end
+    gIX_ = ceil((1:length(cIX_))'/length(cIX_)*min(20,length(cIX_)));
+    UpdateIndices(hfig,cIX_,gIX_,40);
+    RefreshFigure(hfig);
+end
+end
+
+%% row 3: (TBD)
+
+function checkbox_dataFR_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+SetDataFR(hfig,get(hObject,'Value'));
+RefreshFigure(hfig);
+end
+
+%% ----- tab four ----- (Clustering etc.)
+
+%% row 1: k-means
+
+function edit_kmeans_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+M = getappdata(hfig,'M');
+
+str = get(hObject,'String');
+if ~isempty(str),
+    temp = textscan(str,'%d');
+    numK = temp{:};
+    disp(['k-means k=' num2str(numK) '...']);
+    tic
+    rng('default');% default = 0, but can try different seeds if doesn't converge
+    if numel(M)*numK < 10^7 && numK~=1,
+        disp('Replicates = 5');
+        [gIX,C,sumd,D] = kmeans(M,numK,'distance','correlation','Replicates',5);
+    elseif numel(M)*numK < 10^8 && numK~=1,
+        disp('Replicates = 3');
+        [gIX,C,sumd,D] = kmeans(M,numK,'distance','correlation','Replicates',3);
+    else
+        [gIX,C,sumd,D] = kmeans(M,numK,'distance','correlation');%,'Replicates',3);
+    end
+    toc
+    beep
+    
+    if numK>1,
+        gIX = HierClusDirect(C,gIX,numK);
+    end
+    
+    cIX = getappdata(hfig,'cIX');
+    UpdateIndices(hfig,cIX,gIX,numK);
+    RefreshFigure(hfig);
+end
+end
+
+function edit_kmeans2_Callback(hObject,~) % based on anatomical distance
+hfig = getParentFigure(hObject);
+z_res = getappdata(hfig,'z_res');
+cIX = getappdata(hfig,'cIX');
+CInfo = getappdata(hfig,'CInfo');
+M = getappdata(hfig,'M');
+
+XY = reshape([CInfo(cIX).center],2,[])';
+Z = [CInfo(cIX).slice]'.*z_res;
+XYZ = horzcat(XY,Z);
+Combo = horzcat(XYZ/50,M);% how to weight????????????????
+
+str = get(hObject,'String');
+if ~isempty(str),
+    temp = textscan(str,'%d');
+    numK = temp{:};
+    disp(['anat. weighted k-means k=' num2str(numK) '...']);
+    tic
+    rng('default');% default = 0, but can try different seeds if doesn't converge
+    [gIX,C,sumd,D] = kmeans(Combo,numK,'distance','correlation');%,'Replicates',3);
+    toc
+    beep
+    
+    if numK>1,
+        gIX = HierClusDirect(C,gIX,numK);
+    end
+
+    cIX = getappdata(hfig,'cIX');
+    UpdateIndices(hfig,cIX,gIX,numK);
+    RefreshFigure(hfig);
+end
+end
+
+function edit_kmeans_elbow_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+gIX = getappdata(hfig,'gIX');
+M = getappdata(hfig,'M');
+
+str = get(hObject,'String');
+if ~isempty(str),
+    str = strrep(str,'end',num2str(max(gIX)));
+    range = ParseRange(str);
+
+    Silh_mean = zeros(1,length(range));
+    for i = 1:length(range),
+        disp(['k-means k=' num2str(range(i))]);
+        gIX = kmeans(M,range(i),'distance','correlation');        
+        silh = silhouette(M,gIX,'correlation');
+        Silh_mean(i) = mean(silh);
+    end
+    [~,ix] = max(Silh_mean);    
+    numK = range(ix);
+    
+    figure;hold on
+    plot(range,Silh_mean,'o-','color',[0.5 0.5 0.5]);
+    
+    if length(range)~=1,
+        plot([numK,numK],[min(Silh_mean),max(Silh_mean)],'r--');
+        xlim([range(1),range(end)]);
+        ylim([min(Silh_mean),max(Silh_mean)])
+    end
+    xlabel('k = # of clusters')
+    ylabel('silhouette mean')
+    
+    %% perform regular k-means
+    disp(['k-means k=' num2str(numK)]);
+    rng('default');% default = 0, but can try different seeds if doesn't converge
+    if numel(M)*numK < 10^7 && numK~=1,
+        disp('Replicates = 5');
+        [gIX,C,sumd,D] = kmeans(M,numK,'distance','correlation','Replicates',5);
+    elseif numel(M)*numK < 10^8 && numK~=1,
+        disp('Replicates = 3');
+        [gIX,C,sumd,D] = kmeans(M,numK,'distance','correlation','Replicates',3);
+    else
+        [gIX,C,sumd,D] = kmeans(M,numK,'distance','correlation');%,'Replicates',3);
+    end       
+    if numK>1,
+        gIX = HierClusDirect(C,gIX,numK);
+    end    
+    cIX = getappdata(hfig,'cIX');    
+    UpdateIndices(hfig,cIX,gIX,numK);
+    RefreshFigure(hfig);
+    beep;
+end
+end
+
+%% row 2: Auto-clustering
+
+function pushbutton_merge_Callback(hObject,~)
+% disp('merging...');
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+M = getappdata(hfig,'M');
+U = unique(gIX);
+numU = length(U);
+[C,D] = FindCentroid(gIX,M);
+
+thres_merge = getappdata(hfig,'thres_merge');
+
+i = 1;
+while i<numU,    
+    c = corr(C(i,:)',C(i+1,:)');
+    if c > thres_merge,
+        IX = find(gIX == U(i+1));
+        gIX(IX)=U(i);
+        U = unique(gIX);
+        numU = length(U);
+        
+        IX = find(gIX == U(i));
+        M_s = M(IX,:);
+        [~,C1,~,D1] = kmeans(M_s,1,'distance','correlation');
+        C(i,:) = C1;
+        D(i) = mean(D1);
+        C(i+1,:) = [];
+        D(i+1) = [];
+    else
+        i = i+1;
+    end
+end
+
+if numU>1,
+    [gIX, numU] = HierClus(M,gIX);
+end
+
+UpdateIndices(hfig,cIX,gIX,numU);
+RefreshFigure(hfig);
+disp('merging complete');
+end
+
+function edit_mergethres_Callback(hObject,~)
+str = get(hObject,'String');
+temp = textscan(str,'%f',1);
+hfig = getParentFigure(hObject);
+setappdata(hfig,'thres_merge',temp{:});
+end
+
+function pushbutton_iter_split(hObject,~)
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+numU = getappdata(hfig,'numK');
+thres_split = getappdata(hfig,'thres_split');
+M_0 = getappdata(hfig,'M_0_fluo');
+classheader = getappdata(hfig,'classheader');
+
+disp('iter. split all, beep when done...');
+thres_size = 10;
+thres_H = thres_split; 
+% thres_H = [0.2;0.15;0.1;0.05]; % could have more rounds...
+
+% initialization
+I_rest = [];
+% loop
+tic
+for round = 1:length(thres_H),
+    disp(['round ' num2str(round) ', numU = ' num2str(numU)]);
+    dthres = 1-thres_H(round);
+    gIX_last = gIX;
+    I_clean_last = cIX;
+    cIX = [];
+    gIX = [];
+    for i = 1:numU,
+        disp(['i = ' num2str(i)]);
+        ix = find(gIX_last == i);
+%         IX = ix;
+        IX = I_clean_last(ix);
+        M_s = M_0(IX,:);
+        [I_rest,cIX,gIX,numU] = CleanClus(M_s,IX,I_rest,cIX,gIX,numU,dthres,thres_size);
+%         cIX = I_clean_last(I_clean);
+    end
+
+    [gIX, numU] = SqueezeGroupIX(gIX);
+    SaveClass(hfig,cIX,gIX,classheader,['clean_round' num2str(round)]);
+
+    SaveClass(hfig,I_rest,ones(length(I_rest),1),classheader,...
+       ['rest_round' num2str(round)]);
+end
+toc
+beep
+end
+
+function [I_rest,I_clean,gIX_clean,numU] = CleanClus(M_s,IX,I_rest,I_clean,gIX_clean,numU,dthres,thres_size)
+I_clean_s = [];
+
+% find numK_s for kmeans
+kmax = min(round(size(M_s,1)/thres_size),30);
+% try numK_s = 1
+numK_s = 1;
+rng('default');
+[gIX_s,~,~,D] = kmeans(M_s,numK_s,'distance','correlation');
+Dist = min(D,[],2);
+if mean(Dist)>dthres,
+    % try numK_s = kmax
+    numK_s = kmax;
+    rng('default');
+    [gIX_s,~,~,D] = kmeans(M_s,numK_s,'distance','correlation');
+    Dist = min(D,[],2);
+    if mean(Dist)<dthres, % find in between value for numK_s
+        numK_s = 2;
+        while 1,
+            % kmeans-cluster by numK_s
+            rng('default');
+            [gIX_s,~,~,D] = kmeans(M_s,numK_s,'distance','correlation');
+            Dist = min(D,[],2);
+            if mean(Dist)<dthres,
+                break;
+            end
+            
+            if numK_s < kmax,
+                numK_s = numK_s+1;
+                disp(['numK_s = ' num2str(numK_s)]);
+            else % numK_s = kmax;
+                break;
+            end
+        end
+    else disp(['numK_s = ' num2str(numK_s)]);
+    end
+end
+% have numK_s that makes mean(Dist) < thres, or numK_s = kmax
+
+for i = 1:numK_s,
+    IX_s = find(gIX_s == i);
+    if length(IX_s)>thres_size, % cluster big enough to start
+        dst = Dist(IX_s);
+        if mean(dst) < dthres,
+            I_clean_s = [I_clean_s; IX_s];
+            gIX_clean = [gIX_clean; gIX_s(IX_s)+double(numU)];
+        else
+            ix = find(dst<dthres); % clean
+            if length(ix)>=thres_size, % clean cluster still big enough
+                I = IX_s(ix);
+                I_clean_s = [I_clean_s; I];
+                gIX_clean = [gIX_clean; gIX_s(I)+double(numU)];
+            end
+        end
+    end
+end
+
+numU = numU + numK_s;
+I_rest_s = setdiff(1:size(M_s,1),I_clean_s);
+I_rest = [I_rest; IX(I_rest_s)];
+I_clean = [I_clean; IX(I_clean_s)];
+end
+
+function edit_splitthres_Callback(hObject,~)
+str = get(hObject,'String');
+temp = textscan(str,'%f',1);
+hfig = getParentFigure(hObject);
+setappdata(hfig,'thres_split',temp{:});
+end
+
+function pushbutton_autoclus_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+M = getappdata(hfig,'M');
+M_0 = getappdata(hfig,'M_0_fluo');
+classheader = getappdata(hfig,'classheader');
+thres_size = 10;
+thres_split = getappdata(hfig,'thres_split');
+thres_stimlock = 1.0;
+thres_merge = getappdata(hfig,'thres_merge');
+thres_silh = 0.4;
+
+isWkmeans = getappdata(hfig,'isWkmeans');
+
+%% kmeans
+if isWkmeans,
+    numK = 20;
+    disp(['kmeans k = ' num2str(numK)]);
+    tic
+    rng('default');% default = 0, but can try different seeds if doesn't converge
+    if numel(M)*numK < 10^7 && numK~=1,
+        disp('Replicates = 5');
+        gIX = kmeans(M,numK,'distance','correlation','Replicates',5);
+    elseif numel(M)*numK < 10^8 && numK~=1,
+        disp('Replicates = 3');
+        gIX = kmeans(M,numK,'distance','correlation','Replicates',3);
+    else
+        gIX = kmeans(M,numK,'distance','correlation');
+    end
+    toc
+end
+[gIX, numU] = SqueezeGroupIX(gIX);
+
+%% pushbutton_iter_split(hObject,~);
+disp('iter. split all...');
+I_rest = [];
+iter = 1;
+gIX_last = gIX;
+I_clean_last = cIX;
+cIX = [];
+gIX = [];
+for i = 1:numU,
+    disp(['i = ' num2str(i)]);
+    ix = gIX_last == i;
+    IX = I_clean_last(ix);
+    M_s = M_0(IX,:);
+    [I_rest,cIX,gIX,numU] = CleanClus(M_s,IX,I_rest,cIX,gIX,numU,1-thres_split,thres_size);
+end
+[gIX, ~] = SqueezeGroupIX(gIX);
+if isempty(gIX),
+    errordlg('nothing to display!');
+    return;
+end
+SaveClass(hfig,cIX,gIX,classheader,['clean_round' num2str(iter)]);
+SaveClass(hfig,I_rest,ones(length(I_rest),1),classheader,...
+    ['rest_round' num2str(iter)]);
+
+[gIX, numU] = Merge_direct(thres_merge,M_0,cIX,gIX);
+
+%% rank by stim-lock
+disp('stim-lock');
+M = M_0(cIX,:);
+[gIX,rankscore] = RankByStimLock_Direct(hfig,cIX,gIX,M,numU);
+disp('ranking complete');
+% and threshold
+IX = find(rankscore<thres_stimlock);
+ix = ismember(gIX,IX);
+gIX = gIX(ix);
+cIX = cIX(ix);
+
+%% ~ pushbutton_autoregclus_Callback(hObject,~);
+thres_reg = getappdata(hfig,'thres_reg');
+[cIX,gIX,~] = AllCentroidRegression_Direct(M_0,thres_reg,cIX,gIX);
+disp('auto-reg-clus complete');
+
+[gIX, numU] = Merge_direct(thres_merge,M_0,cIX,gIX);
+
+SaveClass(hfig,cIX,gIX,classheader,'clean_round2');
+
+%% Silhouette
+disp('silhouette analysis');
+gIX_last = gIX;
+for i = 1:numU,
+    disp(['i = ' num2str(i)]);
+    IX = find(gIX_last == i);
+    cIX_2 = cIX(IX);
+    M_s = M_0(cIX_2,:);
+    % try k-means with k=2, see whether to keep
+    gIX_ = kmeans(M_s,2,'distance','correlation');
+    silh = silhouette(M_s,gIX_,'correlation');
+    if mean(silh)>thres_silh,
+        % keep the k-means k=2 subsplit
+        disp('split');
+        gIX(IX) = gIX_ + numU; % reassign (much larger) gIX
+    end
+end
+[gIX, ~] = SqueezeGroupIX(gIX);
+
+SaveClass(hfig,cIX,gIX,classheader,'clean_round3');
+
+%% rank by stim-lock ?? bug?
+% disp('stim-lock');
+% M = M_0(cIX,:);
+% [gIX,rankscore] = RankByStimLock_Direct(hfig,cIX,gIX,M,numU);
+% disp('ranking complete');
+% % and threshold
+% IX = find(rankscore<thres_stimlock);
+% ix = ismember(gIX,IX);
+% gIX = gIX(ix);
+% cIX = cIX(ix);
+% 
+% [gIX, ~] = Merge_direct(thres_merge,M_0,cIX,gIX);
+
+% size threshold
+thres_size = getappdata(hfig,'thres_size');
+[cIX, gIX, numU] = ThresSize(cIX,gIX,thres_size);
+
+%% update GUI
+if isempty(gIX),
+    errordlg('nothing to display!');
+    return;
+end
+UpdateIndices(hfig,cIX,gIX,numU);
+RefreshFigure(hfig);
+beep;
+
+end
+
+function pushbutton_thressize_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+thres_size = getappdata(hfig,'thres_size');
+[cIX, gIX, numU] = ThresSize(cIX,gIX,thres_size);
+UpdateIndices(hfig,cIX,gIX,numU);
+RefreshFigure(hfig);
+end
+
+function edit_sizethres_Callback(hObject,~)
+str = get(hObject,'String');
+if ~isempty(str),
+    temp = textscan(str,'%f');
+    thres_size = temp{:};
+end
+hfig = getParentFigure(hObject);
+setappdata(hfig,'thres_size',thres_size);
+end
+
+function [cIX, gIX, numU] = ThresSize(cIX,gIX,thres_size)
+U = unique(gIX);
+numU = length(U);
+for i=1:numU,
+    if length(find(gIX==U(i)))<thres_size,
+        cIX(gIX==U(i)) = [];
+        gIX(gIX==U(i)) = [];
+    end
+end
+[gIX, numU] = SqueezeGroupIX(gIX);
+end
+
+function [gIX, numU] = Merge_direct(thres_merge,M_0,cIX,gIX)
+M = M_0(cIX,:);
+[gIX, numU] = HierClus(M,gIX);
+U = unique(gIX);
+M = M_0(cIX,:);
+[C,D] = FindCentroid(gIX,M);
+i = 1;
+while i<numU,    
+    c = corr(C(i,:)',C(i+1,:)');
+    if c > thres_merge,
+        IX = find(gIX == U(i+1));
+        gIX(IX)=U(i);
+        U = unique(gIX);
+        numU = length(U);
+        
+        IX = find(gIX == U(i));
+        M_s = M(IX,:);
+        [~,C1,~,D1] = kmeans(M_s,1,'distance','correlation');
+        C(i,:) = C1;
+        D(i) = mean(D1);
+        C(i+1,:) = [];
+        D(i+1) = [];
+    else
+        i = i+1;
+    end
+end
+[gIX, numU] = HierClus(M,gIX);
+disp('merging complete');
+end
+
+function checkbox_wkmeans_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+setappdata(hfig,'isWkmeans',get(hObject,'Value'));
+end
+
+%% row 3: misc plots
+
+function pushbutton_hierplot_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+M = getappdata(hfig,'M');
+
+[gIX2, numU] = HierClus(M,gIX,'isplotfig');
+if ~isequal(gIX,gIX2),
+    UpdateIndices(hfig,cIX,gIX2,numU);
+    RefreshFigure(hfig);
+end
+end
+
+function edit_hierpartn_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+M = getappdata(hfig,'M');
+gIX = getappdata(hfig,'gIX');
+hierinplace = getappdata(hfig,'hierinplace');
+
+[gIX, numU] = SqueezeGroupIX(gIX);
+[C,~] = FindCentroid(gIX,M);
+
+str = get(hObject,'String');
+if ~isempty(str),
+    temp = textscan(str,'%d',1);
+    numCuts = temp{:};
+    
+%     tree = linkage(C,'average','correlation');
+%     IX_tree = cluster(tree,'maxclust',numCuts);
+    IX_tree = clusterdata(C,'criterion','distance','distance','correlation','maxclust',numCuts)
+    
+    if hierinplace,
+        % sort to keep clusters in place
+        U = unique(IX_tree,'stable');
+        temp = zeros(size(IX_tree));
+        for i = 1:length(U),
+            temp(IX_tree==U(i)) = i;
+        end
+        IX_tree = temp;
+    end
+    
+    % update gIX
+    temp = zeros(size(gIX));
+    for i = 1:numU,
+        temp(gIX==i) = IX_tree(i);
+    end
+    gIX = temp;
+
+    cIX = getappdata(hfig,'cIX');
+    UpdateIndices(hfig,cIX,gIX,length(unique(IX_tree)));
+    RefreshFigure(hfig);
+end
+end
+
+function edit_hierpartthres_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+M = getappdata(hfig,'M');
+gIX = getappdata(hfig,'gIX');
+hierinplace = getappdata(hfig,'hierinplace');
+
+[gIX, numU] = SqueezeGroupIX(gIX);
+[C,~] = FindCentroid(gIX,M);
+
+str = get(hObject,'String');
+if ~isempty(str),
+    temp = textscan(str,'%f',1);
+    thres= temp{:};
+    
+    IX_tree = clusterdata(C,'criterion','distance',...
+        'distance','correlation','cutoff',thres);
+        
+    if hierinplace,
+        % sort to keep clusters in place
+        U = unique(IX_tree,'stable');
+        temp = zeros(size(IX_tree));
+        for i = 1:length(U),
+            temp(IX_tree==U(i)) = i;
+        end
+        IX_tree = temp;
+    end
+    
+    % update gIX
+    temp = zeros(size(gIX));
+    for i = 1:numU,
+        temp(gIX==i) = IX_tree(i);
+    end
+    gIX = temp;
+
+    cIX = getappdata(hfig,'cIX');
+    UpdateIndices(hfig,cIX,gIX,length(unique(IX_tree)));
+    RefreshFigure(hfig);
+end
+end
+
+function checkbox_hierinplace_Callback(hObject,~)
+hierinplace = get(hObject,'Value');
+hfig = getParentFigure(hObject);
+setappdata(hfig,'hierinplace',hierinplace);
+end
+
+function pushbutton_corrplot_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+gIX = getappdata(hfig,'gIX');
+M = getappdata(hfig,'M');
+[C,~] = FindCentroid(gIX,M);
+coeffs = corr(C');%corr(C(1,:)',C(2,:)')
+CorrPlot(coeffs);
+end
+
+function CorrPlot(coeffs)
+im = coeffs;
+
+% red-white-blue colormap
+cmap = zeros(64,3);
+cmap(:,1) = [linspace(0,1,32), linspace(1,1,32)];
+cmap(:,2) = [linspace(0,1,32), linspace(1,0,32)];
+cmap(:,3) = [linspace(1,1,32), linspace(1,0,32)];
+minlim = -1; %min(min(im));
+maxlim = 1; %max(max(im));
+
+RGB = ImageToRGB(im,cmap,minlim,maxlim); % map image matrix to range of colormap
+
+figure('Position',[1000,200,500,500]);
+image(RGB); axis equal; axis tight;
+
+for i = 1:size(im,2), % horizontal.. because of image axis
+    for j = 1:size(im,1),
+        text(i-0.3, j, num2str(round(im(i,j)*100)/100));%, 'Units', 'data')
+    end
+end
+end
+
+function RGB = ImageToRGB(im,cmap,minlim,maxlim)
+L = size(cmap,1);
+ix = round(interp1(linspace(minlim,maxlim,L),1:L,im,'linear','extrap'));
+RGB = reshape(cmap(ix,:),[size(ix) 3]); % Make RGB image from scaled.
+end
+
+function pushbutton_plot4x4_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+PlotBy16StimsFromGUI(hfig);
+end
+
+%% ----- tab five ----- (Saved Clusters)
+
+%% row 1: Class
+
+function popup_classmenu_Callback(hObject,~)
+classID = get(hObject,'Value') - 1;
+if classID>0,
+    hfig = getParentFigure(hObject);
+    UpdateClassID(hfig,classID); % update popupmenu
+end
+end
+
+function edit_editclassname_Callback(hObject,~)
+str = get(hObject,'String');
+
+hfig = getParentFigure(hObject);
+Class = getappdata(hfig,'Class');
+classID = getappdata(hfig,'classID');
+Class(classID).name = str;
+setappdata(hfig,'Class',Class);
+UpdateClassID(hfig,classID); % update popupmenu
+end
+
+function pushbutton_saveclass_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+SaveClass(hfig);
+end
+
+function edit_class_header_Callback(hObject,~)
+% get/format range
+str = get(hObject,'String');
+
+hfig = getParentFigure(hObject);
+setappdata(hfig,'classheader',[str ':']);
+end
+
+function edit_newclassname_Callback(hObject,~)
+str = get(hObject,'String');
+hfig = getParentFigure(hObject);
+setappdata(hfig,'newclassname',str);
+end
+
+function pushbutton_makeclass_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+classheader = getappdata(hfig,'classheader');
+newclassname = getappdata(hfig,'newclassname');
+SaveClass(hfig,cIX,gIX,classheader,newclassname);
+end
+
+function pushbutton_delclass_Callback(hObject,~)
+choice = questdlg('Delete current class?','','Cancel','Yes','Yes');
+if strcmp(choice,'Yes'), 
+    hfig = getParentFigure(hObject);
+    Class = getappdata(hfig,'Class');
+    classID = getappdata(hfig,'classID');
+    disp(['delete ' num2str(classID)]);
+    Class(classID) = [];
+    setappdata(hfig,'Class',Class);
+        
+    classID = max(1,classID-1);
+    disp(['new classID: ' num2str(classID)]);
+    UpdateClassID(hfig,classID);
+end
+end
+
+%% row 2: Cluster
+
+function popup_clusgroupmenu_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+clusgroupID = getappdata(hfig,'clusgroupID');
+new_clusgroupID = get(hObject,'Value');
+UpdateClusGroupID(hfig,clusgroupID,new_clusgroupID);
+end
+
+function popup_clusmenu_Callback(hObject,~)
+clusID = get(hObject,'Value') - 1;
+if clusID>0,
+    hfig = getParentFigure(hObject);
+    UpdateClusID(hfig,clusID);
+end
+end
+
+function edit_editclusname_Callback(hObject,~)
+str = get(hObject,'String');
+
+hfig = getParentFigure(hObject);
+Cluster = getappdata(hfig,'Cluster');
+clusID = getappdata(hfig,'clusID');
+Cluster(clusID).name = str;
+setappdata(hfig,'Cluster',Cluster);
+UpdateClusID(hfig,clusID); % update popupmenu
+end
+
+function pushbutton_saveclus_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+SaveCluster(hfig,'current');
+end
+
+function edit_clus_header_Callback(hObject,~)
+% get/format range
+str = get(hObject,'String');
+
+hfig = getParentFigure(hObject);
+setappdata(hfig,'clusheader',[str ':']);
+end
+
+function edit_newclusname_Callback(hObject,~)
+str = get(hObject,'String');
+hfig = getParentFigure(hObject);
+setappdata(hfig,'newclusname',str);
+end
+
+function pushbutton_makeclus_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+SaveCluster(hfig,'new');
+end
+
+function edit_setrank_Callback(hObject,~)
+str = get(hObject,'String');
+C = textscan(str,'%d');
+rank = C{:};
+
+hfig = getParentFigure(hObject);
+Cluster = getappdata(hfig,'Cluster');
+clusID = getappdata(hfig,'clusID');
+% insert current cluster into new position = 'rank'
+if rank > 1,
+    temp = Cluster(clusID);
+    Cluster(clusID) = [];
+    Cluster = [Cluster(1:rank-1),temp,Cluster(rank:end)];
+else 
+    temp = Cluster(clusID);
+    Cluster(clusID) = [];
+    Cluster = [temp,Cluster(rank:end)];
+end
+setappdata(hfig,'Cluster',Cluster);
+% update pointer = clusID
+clusID = rank;
+UpdateClusID(hfig,clusID);
+end
+
+function edit_notes_Callback(hObject,~)
+str = get(hObject,'String');
+hfig = getParentFigure(hObject);
+Cluster = getappdata(hfig,'Cluster');
+clusID = getappdata(hfig,'clusID');
+Cluster(clusID).notes = str;
+setappdata(hfig,'Cluster',Cluster);
+end
+
+function pushbutton_delclus_Callback(hObject,~)
+choice = questdlg('Delete current cluster?','','Cancel','Yes','Yes');
+if strcmp(choice,'Yes'), 
+    hfig = getParentFigure(hObject);
+    Cluster = getappdata(hfig,'Cluster');
+    clusID = getappdata(hfig,'clusID');
+    disp(['delete ' num2str(clusID)]);
+    Cluster(clusID) = [];
+    setappdata(hfig,'Cluster',Cluster);
+        
+    clusID = max(1,clusID-1);
+    disp(['new clusID: ' num2str(clusID)]);
+    UpdateClusID(hfig,clusID);
+end
+end
+
+%% row 3: misc
+
+function edit_clusUnion_Callback(hObject,~)
+disp('union processing...');
+hfig = getParentFigure(hObject);
+M_0 = GetM_0(hfig);
+Cluster = getappdata(hfig,'Cluster');
+str = get(hObject,'String');
+if ~isempty(str),
+    % get/format range
+    str = strrep(str,':','-'); % e.g. str= '1,3,5:8';
+    C = textscan(str,'%d','delimiter',',');
+    m = C{:};
+    range = [];
+    for i = 1:length(m),
+        if m(i)>0,
+            range = [range,m(i)]; %#ok<AGROW>
+        else % have '-'sign,
+            range = [range,m(i-1)+1:-m(i)]; %#ok<AGROW>
+        end
+    end
+
+    % combine
+    CIX = vertcat(Cluster(range).cIX);
+    GIX = []; % gIX to match A
+    for i = 1:length(range),
+        GIX = [GIX; ones(length(Cluster(range(i)).gIX),1)*i];
+    end    
+    [cIX,gIX,numK] = SmartUnique(CIX,GIX,M_0(CIX,:));    
+    UpdateIndices(hfig,cIX,gIX,numK);    
+    RefreshFigure(hfig);
+end
+disp('union complete');
+end
+
+function [cIX,gIX,numK] = SmartUnique(CIX,GIX,M) % input: simply concatenated groups
+disp('unique based on corr coeff...');
+CTRD = FindCentroid(GIX,M); 
+
+[uCIX,ia,~] = unique(CIX);
+uGIX = GIX(ia);
+if length(uCIX) == 1, 
+    counts = length(x);
+else counts = hist(CIX,uCIX); 
+end
+IX1 = find(counts==1);
+CIX1 = uCIX(IX1); % single occurence
+GIX1 = uGIX(IX1);
+ia = find(~ismember(CIX,CIX1));
+C = CIX(ia); % multiple copies/occurence, keep all copies
+G = GIX(ia);
+M_s = M(ia,:);
+Coeff = zeros(length(ia),1);
+for i = 1:length(ia),
+    reg = CTRD(G(i),:);
+    Coeff(i) = corr(M_s(i,:)',reg');
+end
+[~,I] = sort(Coeff,'descend'); % rank by coeff, so unique ('stable') gets highest
+C = C(I); % finish sorting
+G = G(I);
+[CIX2,ia,~] = unique(C,'stable'); % keep the order!
+GIX2 = G(ia); % the chosen copy for those with multiple copies
+cIX = vertcat(CIX1,CIX2);
+gIX = vertcat(GIX1,GIX2);
+[gIX, numK] = SqueezeGroupIX(gIX);
+disp('smart union complete');
+end
+
+function pushbutton_newclusgroup_Callback(hObject,~)
+hfig = getParentFigure(hObject);
+ClusGroup = getappdata(hfig,'ClusGroup');
+ClusGroup = [ClusGroup,cell(1,1)];
+setappdata(hfig,'ClusGroup',ClusGroup);
+
+clusgroupID = getappdata(hfig,'clusgroupID');
+new_clusgroupID = length(ClusGroup);
+UpdateClusGroupID(hfig,clusgroupID,new_clusgroupID);
+end
+
+function pushbutton_delclusgroup_Callback(hObject,~)
+choice = questdlg('Delete current Cluster-group?','','Cancel','Yes','Yes');
+if strcmp(choice,'Yes'),
+    hfig = getParentFigure(hObject);
+    ClusGroup = getappdata(hfig,'ClusGroup');
+    clusgroupID = getappdata(hfig,'clusgroupID');
+    ClusGroup(clusgroupID) = [];
+    setappdata(hfig,'ClusGroup',ClusGroup);
+    UpdateClusGroupID(hfig,[],max(1,clusgroupID-1));
+end
+end
+
+%% Internal functions
+
+function SetDataFR(hfig,dataFR)
+cIX = getappdata(hfig,'cIX');
+setappdata(hfig,'dataFR',dataFR);
+if dataFR,
+    M_0_fluo = getappdata(hfig,'M_0_fluo');
+    setappdata(hfig,'M',M_0_fluo(cIX,:));
+else % Unchecked
+    M_0_reg = getappdata(hfig,'M_0_reg');
+    setappdata(hfig,'M',M_0_reg(cIX,:));
+end
+global hdataFR;
+set(hdataFR,'Value',dataFR);
+end
+
+function SaveClass(hfig,cIX,gIX,classheader,newclassname)
+if ~exist('cIX','var'),
+    cIX = getappdata(hfig,'cIX');
+end
+if ~exist('gIX','var'),
+    gIX = getappdata(hfig,'gIX');
+end
+
+Class = getappdata(hfig,'Class');
+dataname = getappdata(hfig,'dataname');
+
+if ~exist('newclassname','var'),
+    classID = getappdata(hfig,'classID');
+else
+    classID = numel(Class)+1;
+    Class(classID).name = [classheader newclassname];
+end
+
+Class(classID).cIX = cIX;
+Class(classID).gIX = gIX;
+Class(classID).numel = length(cIX);
+U = unique(gIX);
+numU = length(U);
+Class(classID).numK = numU;
+Class(classID).dataname = dataname;
+
+setappdata(hfig,'Class',Class);
+UpdateClassID(hfig,classID); 
+disp('class saved');
+end
+
+function UpdateClassID(hfig,classID,norefresh)
+Class = getappdata(hfig,'Class');
+numK = Class(classID).numK;
+i_fish = getappdata(hfig,'i_fish');
+% save
+setappdata(hfig,'classID',classID);
+% update GUI
+global hclassname hclassmenu;
+if ishandle(hclassname),
+    set(hclassname,'String',Class(classID).name);
+    temp = MakeMenu({Class.name});
+    set(hclassmenu, 'String', temp);
+    set(hclassmenu, 'Value', classID+1);
+end
+
+if ~exist('norefresh','var'),
+    UpdateIndices(hfig,Class(classID).cIX,Class(classID).gIX,numK);
+    RefreshFigure(hfig);
+end
+global VAR;
+VAR(i_fish).Class = getappdata(hfig,'Class');
+end
+
+function UpdateClusGroupID(hfig,clusgroupID,new_clusgroupID,norefresh)
+% save/update old Cluster into ClusGroup before exiting, 
+% as Cluster is the variable handled in hfig but not saved elsewhere
+ClusGroup = getappdata(hfig,'ClusGroup');
+Cluster = getappdata(hfig,'Cluster');
+i_fish = getappdata(hfig,'i_fish');
+
+ClusGroup{clusgroupID} = Cluster;
+setappdata(hfig,'ClusGroup',ClusGroup);
+global VAR;
+VAR(i_fish).ClusGroup = CurrentClusGroup(hfig);
+
+Cluster = ClusGroup{new_clusgroupID};
+setappdata(hfig,'Cluster',Cluster);
+setappdata(hfig,'clusgroupID',new_clusgroupID);
+
+global hclusgroupmenu;
+if ishandle(hclusgroupmenu),
+    num = length(ClusGroup);
+    menu = {}; for i = 1:num, menu = [menu,{num2str(i)}];end
+    set(hclusgroupmenu,'String', menu,'Value',new_clusgroupID);
+end
+
+if ~exist('norefresh','var'),    
+    if numel(Cluster) == 0,
+        SaveCluster(hfig,'new');
+    else
+        clusID = 1;
+        UpdateClusID(hfig,clusID);
+    end
+end
+end
+
+function ClusGroup = CurrentClusGroup(hfig)
+ClusGroup = getappdata(hfig,'ClusGroup');
+Cluster = getappdata(hfig,'Cluster');
+clusgroupID = getappdata(hfig,'clusgroupID');
+ClusGroup{clusgroupID} = Cluster;
+setappdata(hfig,'ClusGroup',ClusGroup);
+end
+
+function [Cluster,clusID] = SaveCluster(hfig,state,clusheader,name)
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+
+Cluster = getappdata(hfig,'Cluster');
+dataname = getappdata(hfig,'dataname');
+
+if strcmp(state,'current'),
+    clusID = getappdata(hfig,'clusID');
+else %if strcmp(state,'new'),
+    clusID = numel(Cluster)+1;
+    if ~exist('name','var'),
+        name = getappdata(hfig,'newclusname');
+    end
+    if ~exist('clusheader','var'),
+        clusheader = getappdata(hfig,'clusheader');
+    end
+    Cluster(clusID).name = [clusheader name];
+end
+
+Cluster(clusID).cIX = cIX;
+Cluster(clusID).gIX = gIX;
+Cluster(clusID).numel = length(cIX);
+U = unique(gIX);
+numU = length(U);
+Cluster(clusID).numK = numU;
+Cluster(clusID).dataname = dataname;
+
+setappdata(hfig,'Cluster',Cluster);
+UpdateClusID(hfig,clusID); 
+disp('cluster saved');
+end
+
+function UpdateClusID(hfig,clusID)
+Cluster = getappdata(hfig,'Cluster');
+    
+% save
+setappdata(hfig,'clusID',clusID);
+% update GUI
+global hclusname hclusmenu;
+set(hclusname,'String',Cluster(clusID).name);
+menu = MakeMenu({Cluster.name});
+set(hclusmenu,'String', menu,'Value',clusID+1);
+numK = Cluster(clusID).numK;
+
+UpdateIndices(hfig,Cluster(clusID).cIX,Cluster(clusID).gIX,numK);
+RefreshFigure(hfig);
+end
+
+function menu = MakeMenu(name) % careful: input {Class.name}, otherwise only get 1 somehow
+menu = [{'(choose)'},name];
+for j=2:length(menu),menu(j)={[num2str(j-1) ': ' menu{j}]};end
+end
+
+function [gIX, numU] = HierClus(M,gIX,isplotfig) %#ok<INUSD>
+[gIX, numU] = SqueezeGroupIX(gIX);
+[C,~] = FindCentroid(gIX,M);
+D = pdist(C,'correlation');
+tree = linkage(C,'average','correlation');
+leafOrder = optimalleaforder(tree,D);
+
+if numU>1,
+    if exist('isplotfig','var'),
+            figure('Position',[100 400 1300 400]);
+            subplot(1,3,1);
+            CORR = corr(C');
+            imagesc(CORR);axis equal;axis tight
+            
+            subplot(1,3,2);
+            dendrogram(tree,numU,'orientation','right','reorder',leafOrder);
+            set(gca,'YDir','reverse');
+            colormap(jet);
+
+            subplot(1,3,3);
+            C2 = C(leafOrder,:);
+            CORR2 = corr(C2');
+            imagesc(CORR2);axis equal;axis tight
+    end
+    % sort for uniform colorscale
+    temp = zeros(size(gIX));
+    for i = 1:numU,
+        temp(gIX==leafOrder(i)) = i; % = T(i) for clusters segmented from tree
+    end
+    gIX = temp;
+end
+end
+
+function gIX = HierClusDirect(C,gIX,numU)
+D = pdist(C,'correlation');
+tree = linkage(C,'average','correlation');
+leafOrder = optimalleaforder(tree,D);
+
+% sort for uniform colorscale
+temp = zeros(size(gIX));
+for i = 1:numU,
+    temp(gIX==leafOrder(i)) = i; % = T(i) for clusters segmented from tree
+end
+gIX = temp;
+end
+
+function [gIX, numK] = SqueezeGroupIX(gIX)
+U = unique(gIX);
+numK = length(U);
+for i = 1:numK,
+    old = U(i);
+    gIX(gIX==old) = i;
+end
+end
+
+% frequently used, updates cell-index,group-index,cluster-number. set-operations included in here.
+function UpdateIndices(hfig,cIX,gIX,numK)
+global hback hopID;
+M_0 = GetM_0(hfig);
+if ~exist('gIX','var'),
+    gIX = getappdata(hfig,'gIX');
+end
+if ~exist('cIX','var'),
+    cIX = getappdata(hfig,'cIX');
+end
+
+% update cache
+bC = getappdata(hfig,'bCache');
+cIX_last = getappdata(hfig,'cIX');
+gIX_last = getappdata(hfig,'gIX');
+if ~(isequal(cIX_last,cIX) && isequal(gIX_last,gIX)),
+    bC.cIX = [cIX_last,bC.cIX];
+    bC.gIX = [gIX_last,bC.gIX];
+    bC.numK = [getappdata(hfig,'numK'),bC.numK];
+    set(hback,'enable','on');
+    if length(bC.cIX)>20,
+        bC.cIX(end) = [];
+        bC.gIX(end) = [];
+        bC.numK(end) = [];
+    end
+end
+
+% set operations, if applicable
+opID = getappdata(hfig,'opID');
+if opID ~= 0,
+    switch opID,
+        case 1,
+            disp('union');
+            [~,ia,ib] = union(cIX_last,cIX,'stable');
+            IX = vertcat(cIX_last(ia),cIX(ib));% IX;
+        case 2,
+            disp('intersect');
+            [IX,ia,~] = intersect(cIX_last,cIX);
+            ib = [];
+        case 3,
+            disp('setdiff');
+            [IX,ia] = setdiff(cIX_last,cIX);
+            ib = [];
+        case 4,
+            disp('rev setdiff');
+            % swap sequence, then same as opID==3
+            temp = cIX;
+            cIX = cIX_last;
+            cIX_last = temp;
+            temp = gIX;
+            gIX = gIX_last;
+            gIX_last = temp;
+            [IX,ia] = setdiff(cIX_last,cIX);
+            ib = [];
+        case 5,
+            disp('setxor');
+            [IX,ia,ib] = setxor(cIX_last,cIX);
+        case 6,
+            disp('smartUnion');
+            CIX = vertcat(cIX_last,cIX);
+            GIX = [gIX_last;gIX+max(gIX_last)]; % gIX to match
+            [cIX,gIX,numK] = SmartUnique(CIX,GIX,M_0(CIX,:));              
+    end
+    if opID<6,
+        if ~isempty(IX),
+            cIX = IX;
+            gIX = vertcat(gIX_last(ia),gIX(ib)+max(gIX_last(ia)));
+            numK = length(unique(gIX));
+            %         [gIX, numK] = SqueezeGroupIX(gIX);
+        else
+            errordlg('operation result is empty set!')
+            waitforbuttonpress;
+        end
+    end
+    set(hopID,'Value',1,'BackgroundColor',[1,1,1]); % reset
+    setappdata(hfig,'opID',0);
+end
+
+if length(cIX)*size(M_0,2)<5*10^8,
+    % set M
+    M = M_0(cIX,:);
+    
+    setappdata(hfig,'M',M);
+    setappdata(hfig,'bCache',bC);
+    setappdata(hfig,'cIX',cIX);
+    setappdata(hfig,'gIX',gIX);
+    if exist('numK','var'),
+        setappdata(hfig,'numK',numK);
+    end
+else
+    errordlg('dataset too large!')
+    waitforbuttonpress;
+end
+
+% handle rankID: >=2 means write numbers as text next to colorbar
+% first UpdateIndices sets rankID to 100, second sets back to 0 
+rankID = getappdata(hfig,'rankID');
+if rankID>=2,
+    if rankID==100,
+        setappdata(hfig,'rankID',0);
+    else
+        setappdata(hfig,'rankID',100);
+    end
+end
+end
+
+% frequently used, 2 plotting functions are outside ('DrawClusters.m' and 'DrawClustersOnMap_LSh.m')
+function RefreshFigure(hfig)
+CInfo = getappdata(hfig,'CInfo');
+M = getappdata(hfig,'M');
+dataFR = getappdata(hfig,'dataFR');
+fictive = getappdata(hfig,'fictive');
+cIX = getappdata(hfig,'cIX');
+gIX = getappdata(hfig,'gIX');
+
+numK = getappdata(hfig,'numK');
+stim = getappdata(hfig,'stim');
+anat_yx = getappdata(hfig,'anat_yx');
+anat_yz = getappdata(hfig,'anat_yz');
+anat_zx = getappdata(hfig,'anat_zx');
+isCentroid = getappdata(hfig,'isCentroid');
+clrmap = getappdata(hfig,'clrmap');
+rankscore = getappdata(hfig,'rankscore');
+rankID = getappdata(hfig,'rankID');
+
+iswrite = (rankID>=2);
+
+if isempty(cIX),
+    errordlg('empty set! go back!');
+    return;
+end
+
+allAxesInFigure = findall(hfig,'type','axes');
+if ~isempty(allAxesInFigure)
+    delete(allAxesInFigure);
+end
+
+figure(hfig);
+% left subplot
+h1 = axes('Position',[0.05, 0.04, 0.55, 0.83]);
+if isCentroid,
+    [C,~] = FindCentroid(gIX,M);
+    DrawClusters(h1,C,unique(gIX),dataFR,numK,stim,fictive,clrmap,rankscore,iswrite);
+else
+    DrawClusters(h1,M,gIX,dataFR,numK,stim,fictive,clrmap,rankscore,iswrite);
+end
+
+% right subplot
+h2 = axes('Position',[0.61, 0.04, 0.35, 0.83]);
+DrawClustersOnMap_LSh(CInfo,cIX,gIX,numK,anat_yx,anat_yz,anat_zx,clrmap);
+end
+
+function M_0 = GetM_0(hfig)
+dataFR = getappdata(hfig,'dataFR');
+if dataFR,
+    M_0 = getappdata(hfig,'M_0_fluo');
+else
+    M_0 = getappdata(hfig,'M_0_reg');
+end
+end
+
+function [C,D] = FindCentroid(gIX,M)
+U = unique(gIX);
+numU = length(U);
+C = zeros(numU,size(M,2));
+D = zeros(numU,1);
+for i=1:numU,
+    IX = find(gIX == U(i));
+    if length(IX)==1,
+        C(i,:) = M(IX,:);
+        D(i) = 1;
+    else
+        M_s = M(IX,:);
+        [~,C1,~,D1] = kmeans(M_s,1,'distance','correlation');
+        C(i,:) = C1;
+        D(i) = mean(D1);
+    end
+end
+end
+
+function closefigure_Callback(hfig,~)
+global EXPORT_autorecover;
+EXPORT_autorecover = getappdata(hfig);
+end
+
+function fig = getParentFigure(fig)
+% if the object is a figure or figure descendent, return the figure. Otherwise return [].
+while ~isempty(fig) && ~strcmp('figure', get(fig,'type'))
+  fig = get(fig,'parent');
+end
+end
+
+function runscript(flag_script,var_script)
+switch flag_script
+    case 'push_cIX_gIX'
+        UpdateIndices(var_script{:});
+        RefreshFigure(var_script{1});        
+end
+end
+% end
diff --git a/first version/GetMotorRegressor.m b/first version/GetMotorRegressor.m
new file mode 100644
index 0000000..73f17e7
--- /dev/null
+++ b/first version/GetMotorRegressor.m	
@@ -0,0 +1,86 @@
+function regressors = GetMotorRegressor(fictive)
+%% generate GCaMP6f kernel
+% GCaMP6f: decay half-time: 400±41; peak: 80±35
+% GCaMP6s: 1796±73, 480±24
+
+fpsec = 1.97;
+
+tlen=size(fictive,2);
+t=0:0.05:8; % sec
+gc6=exp(-(t-(4+0.48))/1.8);
+gc6(t<(4+0.48))=0;
+t_im = 0:1/fpsec:1/fpsec*(tlen-1);
+t_gc6=0:0.05:tlen; % sec
+%%
+%     turn(data.swimStartIndT(i)-100,1)  = turn_directLeft(i);     %binary: left turns
+%     turn(data.swimStartIndT(i)-100,2)  = turn_directRight(i);    %binary: right turns
+%     turn(data.swimStartIndT(i)-100,3)  = turn_direct(i);         %binary: all turns (+1 for left -1 for right)
+%     turn(data.swimStartIndT(i)-100,4)  = forward_direct(i);      %binary: forward swims
+%     turn(data.swimStartIndT(i)-100,5)  = swim_direct(i);         %binary: all swims
+%     
+%     turn(data.swimStartIndT(i)-100,6)  = turn_amp(i);            %weighted: all turns
+%     turn(data.swimStartIndT(i)-100,7)  = turn_left(i);           %weighted: left turns
+%     turn(data.swimStartIndT(i)-100,8)  = turn_right(i);          %weighted: right turns
+%     turn(data.swimStartIndT(i)-100,9)  = forward_amp(i);         %weighted: forward swims
+%     turn(data.swimStartIndT(i)-100,10) = swim_amp(i);            %weighted: all swims
+%     
+%     turn(data.swimStartIndT(i)-100,11) = left_amp(i);            %weighted: left channel % 11,12 similar to 10?!
+%     turn(data.swimStartIndT(i)-100,12) = right_amp(i);           %weighted: right channel 
+%     turn(:,13)  = fltCh1;                                            %analog: left channel   
+%     turn(:,14)  = fltCh2;                                            %analog: right channel   
+regressor_0={ % rows = [7,8,9,13,14];
+    fictive(1,:);   %weighted: right turns
+    fictive(2,:);   %weighted: left turns
+    fictive(3,:);   %weighted: forward swims
+    fictive(4,:);  %analog: right channel
+    fictive(5,:);  %analog: left channel   
+    fictive(4,:)+fictive(5,:);   %analog: average
+    };
+nRegType = length(regressor_0);
+name_array = {'w_right','w_left','w_fwd','raw_right','raw_left','raw_all'};
+
+% segment length and round off, for shuffled control
+segLength = floor(tlen/80);
+s4_ = tlen-mod(tlen,segLength);
+
+% initialize/preallocate struct
+regressors(nRegType).name = [];
+regressors(nRegType).im = [];
+regressors(nRegType).ctrl = [];
+
+%% feed all regressors into struct
+idx = 0;
+for j=1:nRegType, %run_StimRegType_subset,
+    len = size(regressor_0{j},1);
+    for i=1:len, %run_PhotoState_subset,
+        idx = idx + 1;
+        reg = regressor_0{j}(i,:);
+        %         idx = (j-1)*nStimRegType + i;
+        
+        regressors(idx).name = [name_array{j} '_' num2str(i)];
+        % generate regressor in imaging time
+        regressors(idx).im = gen_reg_im(gc6, t_gc6, t_im, reg);
+        
+        % generate shuffled regressor
+        reg_ = reg(1:s4_);
+        reg2D = reshape(reg_, segLength, []);
+        indices = randperm(size(reg2D,2)); % reshuffle by indexing
+        shffreg = reg2D(:, indices);
+        shffreg = reshape(shffreg,1,[]);
+        temp = reg;
+        temp(1:s4_) = shffreg;
+        shffreg = temp;
+        regressors(idx).ctrl = gen_reg_im(gc6, t_gc6, t_im, shffreg);
+        
+    end
+end
+
+end
+
+function reg_im = gen_reg_im(gc6, t_gc6, t_im, reg)
+
+temp1=interp1(t_im,reg,t_gc6,'linear','extrap');
+temp2=conv(temp1,gc6,'same');
+reg_im = interp1(t_gc6,temp2,t_im,'linear','extrap');
+
+end
diff --git a/GetPhotoTrans.m b/first version/GetPhotoTrans.m
similarity index 100%
rename from GetPhotoTrans.m
rename to first version/GetPhotoTrans.m
diff --git a/first version/GetStimBar.m b/first version/GetStimBar.m
new file mode 100644
index 0000000..2f7215d
--- /dev/null
+++ b/first version/GetStimBar.m	
@@ -0,0 +1,25 @@
+function stimbar = GetStimBar(halfbarheight,stim) % horizontal stimulus bar. nFrames is length of actual content, barlength is raw length including padding
+m1 = 0.3*ones(halfbarheight,length(stim));
+m2 = m1; % bottom half
+x = stim;
+
+% 0 = all black; 1 = black/white; 2 = white/black; 3 = all white; 4 = all gray;
+% 5 = gray/black; 6 = white/gray; 7 = black/gray; 8 = gray/white.
+% 10 = forward grating (very slow, more for calibration)
+% 11 = rightward grating
+% 12 = leftward grating
+
+% top (~ projection left)
+m1(:,x==0 | x==2 | x==5) = 0; % black
+m1(:,x==4 | x==6 | x==7) = 0.8; % grey
+m1(:,x==1 | x==3 | x==8) = 1; % white
+% bottom (~ projection right)
+m2(:,x==0 | x==1 | x==7) = 0;
+m2(:,x==4 | x==5 | x==8) = 0.8;
+m2(:,x==2 | x==3 | x==6) = 1;
+% colors for OMR
+stimbar = repmat(vertcat(m1,m2),[1 1 3]); % 3 color layers in dim3
+stimbar(:,x==11,1) = 1; % red ~ left
+stimbar(:,x==10,2) = 1; % green ~ forward
+stimbar(:,x==12,3) = 1; % blue ~ right
+end
diff --git a/first version/GetStimRegressor.m b/first version/GetStimRegressor.m
new file mode 100644
index 0000000..ba2d9d5
--- /dev/null
+++ b/first version/GetStimRegressor.m	
@@ -0,0 +1,186 @@
+%%
+% all photoStates:
+% 0 = all black; 1 = black/white; 2 = white/black; 3 = all white; 4 = all gray;
+% 5 = gray/black; 6 = white/gray; 7 = black/gray; 8 = gray/white.
+% 10 = forward grating (very slow, more for calibration)
+% 11 = rightward grating
+% 12 = leftward grating
+
+%  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15
+% 00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33
+% transition e.g.:
+% photostate: 2     3     1     3     3     0     0     1     1     0     3     2     0     2     2     1
+% phototrans: 6    11    13     7    15    12     0     1     5     4     3    14     8     2    10     9
+
+%%
+function [regressors, names] = GetStimRegressor(stim,fishset)
+
+fpsec = 1.97; % should import from data...
+
+%% generate GCaMP6f kernel
+% GCaMP6f: decay half-time: 400±41; peak: 80±35 (ms)
+halftime = 0.4;
+delay = 0.08;
+% GCaMP6s: 1796±73, 480±24 (ms)
+% halftime = 1.8;
+% delay = 0.48;
+
+% the 2014-15 Janelia dataset used nucleus localized GCaMP6f...
+
+T = 8; % kernel time length in sec
+
+tlen=size(stim,2);
+t=0:0.05:T; % sec
+gc6=exp(-(t-(T/2+delay))/halftime);
+gc6(t<(T/2+delay))=0;
+t_im = 0:1/fpsec:1/fpsec*(tlen-1);
+t_gc6=0:0.05:tlen; % sec
+
+%% 
+if fishset == 1,
+    States = [0,1,2,3];
+    names = {'black','phototaxis left','phototaxis right','white',...
+        'right on','left on','right off','left off'};
+elseif fishset == 2,
+    States = [0,1,2,3,4,10,11,12];
+    names = {'black','phototaxis left','phototaxis right','white','grey','OMR forward','OMR left','OMR right',...
+        'left PT&OMR','right PT&OMR'};
+end
+tlen=length(stim);
+impulse = 6; % in frames % arbiturary at this point
+
+%% single photoStates
+numPS = length(States);
+stimPS_on = zeros(numPS,tlen); % PS = photoState
+% stimPS_start = zeros(numPS,tlen); % binary, ~ impulse1
+% stimPS_stop = zeros(numPS,tlen);
+for i = 1:numPS,
+    if ~isempty(find(stim==States(i),1)),
+        stimPS_on(i,:) = (stim==States(i));
+        
+%         diff_A = [stimPS_on(i,1) diff(stimPS_on(i,:))];
+%         %     if i==8, diff_A(end)=-1; end % manual correction: series always ends with PS=8
+%         ind_start_A = find(diff_A==1);
+%         ind_stop_A = find(diff_A==-1);
+%         
+%         for k = 1:length(ind_start_A),
+%             if (ind_start_A(k)+impulse<tlen) % 1*fcSampleRate = samples per 1 sec
+%                 stimPS_start(i,ind_start_A(k):(ind_start_A(k)+impulse)) = 1;
+%             else
+%                 stimPS_start(i,ind_start_A(k):end) = 1;
+%             end
+%         end
+%         for k = 1:length(ind_stop_A),
+%             if (ind_stop_A(k)+impulse<tlen)
+%                 stimPS_stop(i,ind_stop_A(k):(ind_stop_A(k)+impulse)) = 1;
+%             else
+%                 stimPS_stop(i,ind_stop_A(k):end) = 1;
+%             end
+%         end
+    end
+end
+
+%% Combos
+if fishset == 1,    
+    %% PT combos   
+    % right on: 2 3
+    % left on:  1 3
+    % right off: 0 1
+    % left off:  0 2
+    H = {[2 3],[1 3],[0 1],[0 2]};
+    numCB1 = 4;
+    stimCB_on = zeros(numCB1, tlen);
+    for i = 1:numCB1,
+        for j = 1:length(H{i});
+            ix = H{i}(j)+1;
+            ix2 = find(stimPS_on(ix,:));
+            if ~isempty(ix2),
+                stimCB_on(i,ix2) = 1;
+            end
+        end
+    end
+%     stimCB_start = zeros(numCB1, tlen);
+%     for i = 1:numCB1,
+%         for j = 1:length(H{i});
+%             ix = H{i}(j)+1;
+%             ix2 = find(stimPS_start(ix,:));
+%             if ~isempty(ix2),
+%                 stimCB_start(i,ix2) = 1;
+%             end
+%         end
+%     end    
+    
+elseif fishset == 2,  
+    %% PT-OMR combos
+    % left PT/OMR:  1 11
+    % right PT/OMR: 2 12
+    
+    H = {[1 11],[2 12]};
+    numCB1 = 2;
+    stimCB_on = zeros(numCB1, tlen);
+    for i = 1:numCB1,
+        for j = 1:length(H{i});
+            temp = (stim==H{i}(j));
+            ix = find(temp);
+            if ~isempty(ix),
+                stimCB_on(i,ix) = 1;
+            end
+        end
+    end
+    
+end
+
+%% pool all into cell array
+regressor_0={
+    stimPS_on;
+    stimCB_on;
+    };
+nRegType = length(regressor_0);
+% % name_array = {'stimPS_on','stimPS_start','stimPS_stop','stimTS','stimCB_on',...
+% %     'stimCB_start','stimCB'};
+
+% segment length and round off, for shuffled control
+segLength = floor(tlen/80);
+s4_ = tlen-mod(tlen,segLength);
+
+% initialize/preallocate struct
+n = nRegType*numPS;
+regressors(n).name = [];
+regressors(n).im = [];
+regressors(n).ctrl = [];
+
+%% feed all regressors into struct 'regressor_stim'
+idx = 0;
+for j=1:nRegType, %run_StimRegType_subset,
+    len = size(regressor_0{j},1);
+    for i=1:len, %run_PhotoState_subset,
+        idx = idx + 1;
+        reg = regressor_0{j}(i,:);
+        %         idx = (j-1)*nStimRegType + i;
+        
+        regressors(idx).name = names{idx};
+% %         regressor_stim(idx).name = [name_array{j} '_' num2str(i)];
+        % generate regressor in imaging time
+        regressors(idx).im = gen_reg_im(gc6, t_gc6, t_im, reg);
+        
+        % generate shuffled regressor
+        reg_ = reg(1:s4_);
+        reg2D = reshape(reg_, segLength, []);
+        indices = randperm(size(reg2D,2)); % reshuffle by indexing
+        shffreg = reg2D(:, indices);
+        shffreg = reshape(shffreg,1,[]);
+        temp = reg;
+        temp(1:s4_) = shffreg;
+        shffreg = temp;
+        regressors(idx).ctrl = gen_reg_im(gc6, t_gc6, t_im, shffreg);
+    end
+end
+end
+
+function reg_im = gen_reg_im(gc6, t_gc6, t_im, reg)
+
+temp1=interp1(t_im,reg,t_gc6,'linear','extrap');
+temp2=conv(temp1,gc6,'same');
+reg_im = interp1(t_gc6,temp2,t_im,'linear','extrap');
+
+end
diff --git a/Load_LSh_step1_cleanup.m b/first version/Load_LSh_step1_cleanup.m
similarity index 100%
rename from Load_LSh_step1_cleanup.m
rename to first version/Load_LSh_step1_cleanup.m
diff --git a/Load_LSh_step2_detrend.m b/first version/Load_LSh_step2_detrend.m
similarity index 100%
rename from Load_LSh_step2_detrend.m
rename to first version/Load_LSh_step2_detrend.m
diff --git a/Load_LSh_step3_align.m b/first version/Load_LSh_step3_align.m
similarity index 100%
rename from Load_LSh_step3_align.m
rename to first version/Load_LSh_step3_align.m
diff --git a/Load_LSh_step4_loadGUI.m b/first version/Load_LSh_step4_loadGUI.m
similarity index 100%
rename from Load_LSh_step4_loadGUI.m
rename to first version/Load_LSh_step4_loadGUI.m
diff --git a/Modeling001.m b/first version/Modeling001.m
similarity index 100%
rename from Modeling001.m
rename to first version/Modeling001.m
diff --git a/PlotBy16StimsFromGUI.m b/first version/PlotBy16StimsFromGUI.m
similarity index 100%
rename from PlotBy16StimsFromGUI.m
rename to first version/PlotBy16StimsFromGUI.m
diff --git a/Plotting.m b/first version/Plotting.m
similarity index 100%
rename from Plotting.m
rename to first version/Plotting.m
diff --git a/Plotting_withinM.m b/first version/Plotting_withinM.m
similarity index 100%
rename from Plotting_withinM.m
rename to first version/Plotting_withinM.m
diff --git a/PoolClustersFromAllFish.m b/first version/PoolClustersFromAllFish.m
similarity index 100%
rename from PoolClustersFromAllFish.m
rename to first version/PoolClustersFromAllFish.m
diff --git a/SortMbystim.m b/first version/SortMbystim.m
similarity index 100%
rename from SortMbystim.m
rename to first version/SortMbystim.m
diff --git a/first version/imNormalize99.m b/first version/imNormalize99.m
new file mode 100644
index 0000000..b35ac8e
--- /dev/null
+++ b/first version/imNormalize99.m	
@@ -0,0 +1,11 @@
+function out=imNormalize99(im)
+
+im=double(im);
+temp=sort(im(:),'descend');
+th1=temp(round(length(im(:))/100));
+th2=min(im(:));
+
+out=(im-th2)/(th1-th2);
+out(out>1)=1;
+
+
diff --git a/first version/push_cIX_gIX.m b/first version/push_cIX_gIX.m
new file mode 100644
index 0000000..0663718
--- /dev/null
+++ b/first version/push_cIX_gIX.m	
@@ -0,0 +1,10 @@
+function push_cIX_gIX(hfig,cIX,gIX,numK)
+% push cIX, gIX to hfig
+if ~exist('numK','var')
+    numK=max(gIX);
+end
+
+var_script={hfig,cIX,gIX,numK};
+GUI_FishExplorer('_place_holder_',0,0,0,1,'push_cIX_gIX',var_script);
+
+end
\ No newline at end of file
diff --git a/first version/read_LSstack_fast_float.m b/first version/read_LSstack_fast_float.m
new file mode 100644
index 0000000..b17963a
--- /dev/null
+++ b/first version/read_LSstack_fast_float.m	
@@ -0,0 +1,15 @@
+function [stack,dim]=read_LSstack_fast_float(fpath,dim)
+%% reads whole stack
+
+info=dir(fpath);
+if (isempty(info))
+    error('File Does not exist');
+end
+
+
+stack=read_LSstack_fast_float_mex64(fpath,int32(dim(1:2)));
+
+size(stack)
+
+
+stack=reshape(stack,dim);
diff --git a/fishpic.jpg b/fishpic.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ef40e83648fc2b5cb2a7645747ad264e97279344
GIT binary patch
literal 25805
zcmeEucbpVO)9~IMcSp`q1&$n6n9bn`=j?3GX*P-AY@W@tI~x&B5m8Y<vJym*jBtVy
zB`AW3<cxrTWI;tZ5DyRT>)j>#sL%7hzxVs^>HFPoPj_{7byanBRZsWU9jiM-X)YJb
z#FT=<f<jB7g+i(OnewdG8z#dsZ#YuM>`$Zc1Tt08&y;Lxo`s+k0O<SQbw3P7yd)VZ
zhoDfbKjsYEUHu*5Ae6!)5WPPQq71A^MKFioMV8rJ9y~Ot>$T4}cP+!6gSwj8Dw-;S
zxV*SD6LlFfYNI3LckrEED+V_lm?}>NBS9C5m8F7#P^>&PsA~>+Iml;=p{`{)5Yj)W
zE1d1M%%akkA>pX2jNPBs&q3qR$~g4?bS8_-<vv@+pwVd%jRDaa{pj3sCZn82E4!b%
z4sHl?1EWrNxn3x~-yL`w)b;+jlF4NMB(r}w>VfEdK0k|w!RQAd`o+>A5=-?9#k%E@
zJisG##T-#QLgL|2Sr#v5568(tUAtyGdYJO#aylON6NyIyxn7(O$Q5t}T_G|Ccu0pH
zA_r{;IvwTia5RXKgYh8daY4R_%QFyqRQjloEGUVS0atE-5BtrH6b_J2mS$0BNe$px
z4wMJ5kY`ZWR6nQ7jl~0G*TLE-9wG-qd5|ZY=1ZT%n4{>4&ht=@apnP%T(^&v|2vKW
zy?Ds7zf`%uVl!wYkp;n!{z+Fo5FYeA$!*ZCTm{t4iIJEPBVB`mBC-3?nEhxhBb`&u
z=9V+SBO9pGK<Hm!9}`8cF6_kJ=|5v(&~sRLw)+Sk0sW~t5_JIr{@nbph<FV0uW*E;
z#&9?=I6rG-fj~S)MlmuREkg|KJP|nr_rd;z!RU<#?I8>g#2&~muUP+9tFoFC!J;vj
zF&%LY>Z*6e!ttoX^|)?7QqlY3Do^50(mR;J=H%$gO7MQsBdY$jk=KQLyyRdG?@^=s
zMgP?5{-ly?MUqGD?ic=3yWA`s4Do#s3x;(DzfG?9zyjDKLd>w-3d2H(tAk^x#G7zR
z_!&&XPs)U<xC(|@I)6Y7Q)zH6319%9O(M8ww-h`8)@2)mq=N+zRi-kF0zw^~uS5NO
zi2+TC0iXcXstlr3#2(}!BJf~0vuTciT&lwbR<kMW5%2&>9$2172wP#54u-8}b;cp0
z8}$M30FnVU31w2XK~IzAo}c*8+5u5n)G$m-(`6_E(1COxVK)1}N_s0(kalL9nsV(k
zIfOQeDPv870&A8sLCne2Cv!mV91v;^lKE#(j&jry@Y>8-db~NJP#gXJGoE}%A!>a9
z<j*6qN$7D~Hk%v1APJuRx#w6GDI#-*Oo7}pElUX^aRvhcC&=c9%+K}gw@XbVmd!%>
zp6zCn*8zY8aJCWbcS=k?3}ght>?5pl1oQ@7wqBU6_o!@iy&;G2@j~mLfB+I!Zup>?
zohe{Dz&eOYbsmL86_E)=Y>m<5F-S~IyTn9uB6M92A&_p=YLWY8B6Y%U5qRw(Fm8!H
z;9vlayPrg%oU}bz$*X-Lw%TVURX8j4p#J`Mo<avbr%!S{s$!^&euf&SrIo(CO8lw8
zzkST}@NqCPkQn};K$t;+vge-kc9mHi!2UubL3Liavv3a)+BpDJ=%a%&q5lb_0z3!=
zMY2%P&WAmL6ftC_r}cslGRF7YENb6FgrE#G0CO2tc`&J-<`9MxN~V+POtT|4vnp<7
z<RqU1`zP2O!oO_rsOf){doVd3r!bzIyFzP9W3)1qMp;~C5QuDMo7ZVh(PZMB;CyyQ
zD&e%qJPxM7Yh|Dwr9qHE{Z8aTum4{WKIkO>{U<$Cisb=Y(3G~B**=>=V8+ZQCT3=b
za=IJTN1T#C!j20fISN7MpLGyP8Nkp3j-Uw$2lbe;^Og~a+tOh6FxY;~qL0`aEDvU8
zd2$H<jrKoDdG1NT$WnN}!WhxAV6m7N2eVU~LkLW=kPEDUK%i2>u-XyIX^1#yiSvyF
zD#T=kfhNMR*x_a~=mD$c(X<4X{$YAtDEA8$+FT~2j@Z)qA{||h$|P#KBa~k$K<WK?
ziNc5n*1`fEEYZO(7*XYCiUKG)ER@0OB+Si%>U9DIEaCwpEU>|17Hmxl!mxw}J3Im=
zEKR~59U_5cG1w0yCRiSZLrKIBEBtWOgQW8mDjjeF77AdM6;3CGI#`XujHJ*5YfzYz
z1jbHlg{eufQ0W{3p-yCk^?rdgDGI=bm_X$bv0x)hpx232ut_G!AvEj69@y*<*gax8
zY)J|{Nr@7+$pk@<#0z77K};uM!w!)kt&{3ur(eL*Nh7dZCZKv`LfDf;#IVc><5onO
zl(F&@`bCH#DM#S|4Y4KVDL52Hys$z8M?6T_qlm##E0WYHl`x4SELa(Z6DY9RK!7Q$
zP^MFnaK<ClCDmG(9v0>h+C1ts%w!31kH!phWWuON!-07YAuXx(fUydxI-LY)fJg@G
zNP#dWGQfJ1K%x>kbb6{lmJ|g&hOj_s6{TRKF^5o~5djexsRA8MEQQgiz!(-AbSA67
z><~MXW{JRN6-PW~y1;>o={ie5;I>Kx9;;b^J0$8P;7!0U!8|suAe@xs5QcPELJ$LT
z)Y*N41S%CI9X3HaCe`YkS^+~Ob;2%*fbEw?VHYIe%cN|bn}xtJnKbE1BO<HJ?D57C
znM@Y&;6x4~qGZWv9$yI2#^fT{A3%&@xiJ~=BNo5hmkjz5dsv=Hh60E?rV#1EA;d>h
znBWM3ghh%VOvI5Gs$jy=43dl~a|mUg7#pFBl$b6KBW#Z{1}9_)Rivcqk_I6XR_T%{
zhfs>De4cbjsDf1tT_z*c(bRGe9Tu9xYA4Ll2r<7pnPj?h2!(EsMwDd5g?_)r2D2ey
zB&H!fT!S#q(jq!uKu8m5Egn8c$hK-@Nk}WCCUuA&4vIu7omDTOisUg}TraeUw5UFZ
zP?Q3HC{J(K1JPk&eOfOKi#!g4+$$H0f;xj&pzw)eDg!H}5{l9)qb>y&QkKdX5opCC
zDvAocx{z4pM=`G<KV-3jhGx9Rq*yOAsRbsp*a9q6%FGwLU^CFQpxEy>+Xa|b9F>{r
zUOPveqFHo$XF$yITcTdKK@2&pl9U&gNKk8zLfnfdBx=-3_4+*$6t!Wgpk88k*qHjT
zNP>qkR3BkVA}lPWkH#b^on4b6eG*pK9`hzJDJ*j+yeWfJ8gqnG8C4FUREs*LUb<9j
zft^7;Qz&&MU1BdAmIlMFfPh1l#{F(libs_){O*8&3Q4J9kJum($RtTm5D|)GTG%T^
z#4?!$_2v+U4N{HF9mC~mxk(n5;nB3xDNBZZ8blqGv30&Au1(4XJipPP=gAc@KLbbQ
za-%1JA!du*;Rx{4){q?7b8i}B<`Bx`o{$7_%H_;(h)BB~3YZqwroAbJLKdbYeu={9
ziD2oVQ{jXO7!T7GK@Sm3N41KiEUL!Ih=NUv(hNzlQs|H65ZduHu2efnAwm}@Emo39
zGrdZ$HEu*Wh%)Mkb8((uNe?Id21u%eSV^@}5K}31Nk&FwP@&<J$0y;a+_JR7C(nBh
zsz@wN7pl}Mnl<CjXjm#pL{nt+0kxb(V+c{B8l};_84IFz(-`WEEv+V4435#^S2JK{
zNa(g`1UeQf#8n!Vhn=HPU<@D{3y+-;hS(aPmE#l=X-%BRRT{~th7;y;gvpRrg7PB5
zj8AKj@vTONSL=eQG9lZeC15H`$n$C$A}H*G@Ek%NqK0iEB&gFk1qxJ5=&+Q4>z5^U
zAvr>#N~SIo5xP+=tcTSiJ<XudtJpw~O(wl9B4+xnxIXBV5Hx#IpD{^u2wgNcYykG#
zK*RM0wNoxd13m*5kn==grXfUEq_We5u2u&93EYT8R1SZdW7Nsj20z1Sbfz?NlpQxl
zomyDL&&Mf7wpz<H!68&C*2U=}32GAPgLG*S^@$B`u~Ler)J6+kL!gii)ddV%lbSRs
z#3s7QAvX)@HiwA_nE7<4$i$Xgaul+}o`hLSx2EZS%xn(W;^MFrd|{D5ly0V(>;Y3E
zU=f-e-ay7|F*uzrF;i~wMqG9=hi^&p-BuGdV}(r~69dWHL#sCBH8Lc))osEJW`)C=
z5cqTqjm1hO{W=hw*))NG)@(M|T+X1@j2UckZ%Au)8Ew=+Sj)i8m?jm`F@hN87UvNf
zB!nAFIHN{JJcL2~n2C{2*mYjgDq*ngUU?j2a7FeMJK>T*dIv%#y&<8;VT_~#7HQJq
zC)1ITO5mW=GjXX7%?;Tp70@zLH0-pR=qx~&GeR<`Qm5I;RkK7EZ``Fyu@#}9+~sm|
z^dTbPO6s{*W&&{|VxGs6@w!nlKWt$M+(9`thcFr9`P?i$#I_2g9tAEy!s4jM!9Y|p
zg~5|B2u&;v#|su0kId-vqAoGPvTD2`g@n#>aJ_7*Od#_{a1}?cVg+qEgt(idu*sqd
zJcB3$;RMeo(Wug4TFPg`)ewsv^u;(@b(re%!<fz<5t{u*s-BR`^!^ZGV6oL|KNmw~
za)TnEHJEY;Eo_S%Xr4L5b|?Z&vxO!1s)9<~CYJ|wL3bL%<WW<Qrm#oZNk>Q)bnw}9
zf5@qD>ExVvD8+E&Y={$<1U(FeIIq}YJBG^%g*lwWd`?0ejEFIRii5HuSTHCdFl8jc
z2-yj@hY%~mDTSX&U_nBvAhbkM5p`mTNK~AT(G+x1)Q-n<6e=-}H=4o{I2VRvQgf0M
z6?<Y%EN$Q_kyr-LM5DR@DNoZyT(g4oNEj}=JxMZfCQIqH#nlqFAsRyBzBq@p$MOi{
z92ZY!PumlEGvCK$DH34=72@(a3CIFFRKi%&>=v+ia(9wU2+b-j*ffxeGBJ}f<?u=L
zF}pBDlS&gj4?nHU<fon{7G$S=J_X5(G16SSO6^F~Ge(P=bg&qis9me(QCT!$NEatX
zJer+r(DIc68qH`-k$O2@l`t7ei=G~^m>FuPBR6C^RcN)TePM<<W8*m^3`Qbh_o$O%
zrY!0dsTpRbH{%MEya<y6dDL-)&q9sfj7Dx`kqMubs*SRxT0f*Qi`X7!Fo!S@2Yxdr
z6jsG?9tX8YXqt$f6SqVingo|Cvy&nh!@<RUaZ<yFxx7r$tQ7`$7G(;86e?aSl!-t(
z8egTPqgqSebKnP=3@YSOPz50tq4l$<4mZaHMMG4kn<s$MR>%<I#}jNNL}nnn8-gGz
zD;r_@?F@MUmKgJnxEr#=GF{%s7I<{P-$g|N4=NUT{1SoADo2tY;Ies?LO85Il3_*G
zVHU#4?7vP3v$P_Zrx6NNMzKI=6s3}ONh;=$rg-k$kh7WGe^Qin<<vr$W(7T1C0<l7
z)`#<T4`BYy_mc=81BQ4YcF=i%*#}`kQly7rnI6`u^spNe!TB&;3TFqVdpz>jiE{|y
zq)Y;W7byb!!SDb;)*1gZuntDRXaP$cGDH_v0w@G@0GxIm0?GhK01jZr@(4XDJ=g~j
z%k;98&LonE7$V$Z6``t##4AXdQ#xELKuM`rm9pVHM4$E~L;+n!Z}bz2w3Vqv)Tl1x
z=V&r&uQP8CZBnC<7MIE7VU-;5dTd@fU>#wgnMk!sA&90~kXA(37-=?3c7wo7PdP0C
zzS+j`g2_(Cz_wb#%GA@HAt%d~UsEDnx0FtGGc_Ww6_et!paN&5fud6pKigmmL|oRO
zjj)I03|EM)@I@5nV1$E52qGSfVw^-c%U>{jnNXlZMY$mZH4z~|E9_YYdqqM4l65ge
zVpOm8$~?A|$0kaL?b4)JCDpSGYD8?3<Bo(VBNyt09xrA<(`MWX42giL&F|Z&U2>5p
z<wX5SPj1L^pG_>moqnenW%3A3CYkm`XjE;+D>wK_l@VoY^#Lb{RU>wbnC`OB^&m8N
zhB0#_YR`ui^thezfCqm&d|^NimRL4!3r6COkd5Q^6B>U6!$O22ND!Pf86~I*2+K<u
zP69Q75m8ppggls!vIwr&V*MX{m^Cg3B+~}CIfNOf&E+K=Y9+1`ndy9~-WK+8)l5qY
z<+_b{luh#ZPyi6e3tI#j+sMg8d}@ObwR+q}waeqM2h|BHq%-1#o|9js*dbmDGjkFV
zMucf1OdLdEqFXp7LD(HhnK&V~6k?}^bcKY4#~?MPbTE8U+^NS=r7Ga!h7-wnf(H{G
zDVs&`^AwU?9ASBUahcR?Mr=ABic5%C6bZ{@T#=QCJ861}lYxZ8avP3;?OeY<B(|iL
zRA6{CmSEh-2{<q#)xtKUX`K8_3HmgoK^BYg*eFe@3$tu8R)!%*W3EJ!uQItroCwtl
zB@wHH$Vlv{TNR8jm43QZ!wo8YkjWBAS)vSyf)Gd89A1uxT!te=FhW5o%Ec^Hmx31;
z*|~IwRBQ=L0~RQxu<%%FK|p8ZfHzuZ9KuM_;4_2$B#4=I$O%ZD*4o7uT8L_rLU}0*
z{d~5P$1^55CIW%N!i+%SQkkL}7RC(OV=@~@rZGDGR+#Tir%YTW9*votfAf(;_&2qL
zn65QB)ifgG;?osuAE`8Jxei)JU;-nA&`=Dp!$h?us18p;&hiCgQKd02WhdtKqY*g=
z#M(?h$`zpwO-h<Ef(VUDT7n@<Fq~#;LJTcOLq>$C+2aR}iVG44VJsALm~<LN0&}N5
zkUC=^^Y)OJu$cW+tKIAO#RR~+P>F2}r$n35MU8QZ8qg!>3dN`of`wA>&*mjDZa7Y0
zAh0EZAqL5|8yT#Cn+LlrN+dUAGl(nnJdqRQdkmUX%Ea^fMG`MVmNG^2^TiC}X4DQ;
zh{}YL!jM?YgViB4p|J#TE=_A@2jglh4Qx@6EOriIRIbI+Jg>x}<r-4~B&udgC8QC|
zKRzvGVkqK|e;{cr8u1%=5qla@8!(uZdBpCBMiDkxGB&n3AM3GQ(Wu%dO;UjqYBfc8
z91<r)0Z64x%caHu+e72gIbedx0d=@iDn_MBDhBk@g6iaUi``073zO^wpD8w4#CeOY
zR8pN5woM=OVKf9YQ6&i5>UPES5=9uOAl02_>EI}Ml_X7?nSKGpvMO^OveOctCz58R
zm^4xoLc{_K+mYWt6hoj}xe|g^Y#G&<GJ<-#0FwIs2t)+zS&E>HpA07je5KwV$XHl9
z9-%~Ce3+3PDHM`NV1q27j95`dUTvc3Ae0u#Aq!3j1Hcei=sJ%p5OyG;L^|Y5&=v9f
z4V4+xWeIa^)`*3xRT<SrHqD8#2(R2tb>IoD-7mK1gyJ@`IL?4Kpa!D%M9e9ZHxx7i
z7KeCDPCzOrP%aZqX%RYU_3A}RJL>aU;_ie-5hHmCx8FyQbV-aFW6EM8d*1sp14qG6
zYI$KgOGyQAgb$KS(;fm=t6dtV(<k#p`81azVj-Ap5yLLaV0uF3cX+57u@lxrwBC#l
z5od`3#xqBuAP6b7nXr=MN=rg2S`2YTG6D;YiIOQPqVp>ZE>b~{;8B`19i<u=K{E}r
zND>LYBW~0MX+F7GY-FjD;Y^6@&08BD*yf@-v$Mcwm%EG^m(?xfDU^yN+Y-c;dN>6$
zL3o1+{VtA_aDsUPG#uk{l8Gq7^~bmzP-~Bics9GsmS3cNejQb)ND6>WljPMnK>*<)
zZetj5Osf(z9cc+LYFwAVu3%VEc2dTbn00PB?vW@W0VUO)VB;DIFru`4SYbrL*$B0o
z>R~$)VCEqPcgRPzMM;y>Y*wL>l$k*iZY?X!NSOo%caTlB@^GG+o_4^JL^3LgQ28iO
zS+j+gcIS7AX%Q<K<?}SUuuGHy3k#W&Mfd<jW64C=sX&Aik8^zT6vq*bI|&ojo>dg5
zgHBjga#9E0T(OCy3&<%*Iys@7P=Hp;l}ZIuVPZHv3OHsk8aWP=Q!3*5yrM{2s}p(M
zbib7dnxbqAoxx#2CN7Ad^<jk;xG;g3$!M1HqL3C%vh#ai9J3+8O9jCOa#{{HRl)?Y
zCJJGO#@CoaV0S2*4&e-{TO3T8L@CHmvNKi@$u}7p7z*()!eP>*ZnngQ8Qjc3-r86R
zt&JVjLNI~>bhws>tF3CSiLPN<+^o0)%tNk;#Uk)9m`_2q8>N#bo<<!rLze6u3Tk;?
zyGrDWL`6O!KX1N#FhSRnCPoA&<UU9i(s{JZV3=u?rcgbha2ul595$f(1V%b68DA1J
zC_)g3yL5nRJ><`@6*^%+>M-Q34Tl%;C3sXBus$X!6%G027?0FofDbetFsX1nu81VU
zOh1%_NlYoTsns!&$FJ7<hzx1c=@4L?JTxmF<~dCHO<k~VKkmo)EkS!)&DQZ~I$UYt
z%SpbD<-`3k91Wl@#1KvSEs&MOqZz3QEKtTonjL3ihE$x?(O7vYM?`+Ykn(Zls12e@
zQziusVkd#uu7n*DqlHWNLl}}uDU3YaW#Pn3Ho4W!W2K=$Bti`_MGlpS&m@Fly`4$t
z<~JlcUZ0zYA{v!G<pwXMQI3^J3e!G<pJLmA!Ew7Grk!C*6BdN8W!kinjFPVqd+{ho
z2Kb05Af?%y5E>&g1s=-#HgQ%+t#z}QZZVn&LF^Qh=QYq-1||;}bv{!q!CcH#DuEN0
z48&tGTxQ(Oln_h;iBK7`5QFO!8$rV$Xc+X#@=_MCy)v1A;i3r$7lRcPYr(ptj6hTe
zM%v8@y4!5Wj1ePU#uicCx;$oBtI=3OPBkl<FtWYi<!~TqB3$ykvxG;06$~S8uqBp}
z_^2t2!DCvu;*>-aq9&3#+-W<cii=Ecy<A9wEd;AvOO+YaIidpzHq{f&@Dj{a#E{5Q
z7)fg7bcmi^A|%E%olTIEs61lg+er&<peIz`D9cW$OlX1@V&gH6l5dZ5u|(42iYP)1
z7ZY`=z+|z6>=~~p?-EC&CXSN>NwqS_2$|(6lOzcR?QD)WlB8og8Wb|fsR?S_XC-`A
zcgDlx>me-8Vmk@9R4CC#z!o}AsBlOF^fbR^8sQU?bb#Z=e10h;vWZhBsoot*L}gko
z*Qm6mEPQXsXy+KHiZDaUqf$LVnvop~WFR=@2IdZQA4-H%P=Zv&^D~73S#44#HRQeD
z5paAMEG0t;I?||okk=-}35^%h2-8YApWp^?W`yTZd8r{5;8KJgP6WAWW>lx;s1yDi
z4?|j|HzmW=I31zGF^}FCi)a93>PZ(6RWhKM9btgrDH>GJNI#XS)o@a&D34+F2*qX*
zCqQ-Q_ifmEj#Fl3L#jYR2439oC1#9cfjm*CJm&%=NZ^!*<WgQJ!emL!3@>=EWmB3t
zu-fg=#{_(^P9|yY44ti^6SN$K!8l(|57H7=yHluixO|GJT7u&d8!ei#aP%OqU?|cY
zI@_yAMhzB}8~0;it1QBHdimMqipG1~Va*|Yd}uhV)9Fzu&=MU=#WA(R&E$LSMw3{`
z@vH21mxKlMk`2k+{-7%opb3o(D~({%%_21k1q8bMLT@rq-MmO5%?YY4GzC+_)+QY+
zpeW%qiwN)_wglJ#IX`5SdNovzMV(@DArYVrRPo@9FT)lDqDdFk11x7eZ?S=CbF)~?
zT-Q{G-L2r7s8LSDt><!G5ss4U*5TRaOh}`m$Bb0q!p1#>n`DIDc9IqV6FClygvNn6
zt#CdHa9~V(9Fuavu4RJ35%aPXLtr(@5CNmm#c?YH1`i*+ztu8RC~(xwB9|3XKw#g_
zO~f3Ma4LZ@5h-11VB^-j=OAT;xDM14iK`P1C(mRHu$_>G$|qSg4?Rt%l3u-$!xVsN
z3`J#_EJf8@#AY+lW;I*xH~M*j6mTlRyCDNcl_qmS5qm9Q+0<zw?vNL3y5d@(k8x!j
z3njgt80vI8akCxrI!s`*m+F!R<u*FzqLWN7-<pt1!ZssEZ}#|5x>_M+#d8StTq;A6
zFeQ}kAX5jaOoW`&#9^98Y_r=SZ@{Oeq6wZJ82z-J#|B<jNXB6DVipt%K<T*O$&UK<
z3QK@3)1cz~+X}iirVtqEasd<Ud~+oc0w$<dt;!wDWaMmr!sy~~2^AH@wsM&&<0nWL
z4MT95)}|BSL@u;884#BX(HW8kb$-Z-1jC{Wn8nU`IA!8%!HN+y6X1{;7AK8#F&%_8
zEE&VBO~~aAK}sw)LNdOZfY>IL$!#`2871Tp{wZ?!t49IY{a^&Jlr>@HM6eXo2^5o1
zpeC_fBqbtJtIsXA;O-EVmoj+0!=>5GaXs4$>AYA}Zezk!QbeU;3R5t`g}_k|f@h8U
zNm|IMbfFnNOf#DTEOEl^GZ9uIVkP)`UfAt-1ODbH3@X!@(?%uTC|?(sC^7I7B*Bfa
zOpK^Zr!nvt)U-}Wg6+$ICBzK*bXtwpAEd{lIG2=jB^+R#>=vLmV#2PL=HHryk|08i
zT0pR<jC13ec*@3*E5UY(KjiTHSp@kol1jufP)KM`3keNQH93tu8s?SI<h-OBA-xj5
z&WJH`6ha!{vLhz8(2%fUJ}Zl$;;P3j$m7Q2uOns59!P;LMbemmQ7EE`fyD=E#sx1?
zsjdK*<2U;vHU=4Vdg7sskDiY3Y*DY(AcP2k!{kz65jG5|Gzyo&Or=>+c_c{CVoBT|
zSLE#>ClC>_eZbYxkx{mg$F&$?9VT>>C`~9~1WXu}s|urTHwXjVX)ToTaS1=4E_G6|
zw9=5#d3_O{(aVk~+_5+(?_-<9T9;Q&wE!+AVk)rf1zxd*V-A(UD`U}MgdlJa3Oiw-
z#$F;GU}&P4&B?J5B1MLw<|F)sQ%BNV8oeQ^aOZc4g?N-@h15PMBynd_z>*`ZxKl#P
zNhgmC;Gh8`Eit>~LTbvPmuYkgJ`;5qfom)<v3MM<gsV?0P$yz?O2kM$09WG(Sl5_h
zGwL#F(@97r&Ddl()xsAflngZojq4>=uEWG}MA#aU0WyGzA2IT>u8KA)HQI5h3DVMd
z<}g!a%by}hVRAbiV+e%~w^@#eRf)I<wxpCouysO8rBbmZ=D@r%8f+4wuz~=?how59
z_|KtveuCxH=E>RkS2OSNvX~phBRm3%f^e9t2I~&j&2uU_q&Y~}a_{fkXnbN+5N1M(
zaKZvH5fH6QAq!X{P>8T-1Hc)H^ErHn$p!?S(?bZ#pJ2@$(##)<{PVHN|H<*;?19;T
zJ!|@Zd5Bo0`sdTY!65WkCy8UE-u38g_+K7Cj%AOyLJv;30`P8sjYyOLhr<8qK=PA_
zL<w*r{LxAC?1^o_C6%gN7K)Kr$l;O+2ZMtCIPNUxGx&U%#zzD^gdt+l={z=D05gSL
zM1=4~Fc--kaecDsBWc}7D-1j0*+b~}(K`PJXdi3(FK9JU9NYuI0*}GO;{Z2X$e{7~
z3>Kd$5DNu#I-kyhd2BvS$fUDrJS2ab7m)lH?f<0@y({Fzv&Zj|02Yf4jzq&@aQfKk
zelYAm)&EO-p(~0fTuyN`9Lya@|Mv%%2j=vm9DgK=+@LV98_D9i-2Ip|8m%8jXR-S6
zT`p!nr_;`4v$-xhpW_+`{Y%?N1ZAb1uXg^cIC!v_%>`1Xi5N5v4@jAZa9}zE5wQd;
z4$Klh!2DR#zrq~;UrQuU*<ZB(t8SuyE0jE%Kh-}%@jz;iN#(J_=l8W$ipRhi_Vi;F
zS=MDwof}-llcf(Tz+H#`(W&$p31)0L=Ey46;1~uJB+KB(${vvny2vLWyez<Ddhwr6
z(SK?B_xO`uSLmPaCgf4&p#C07EbJzeSkwi3z^wnf`vG}OPlEnEN{1I*Byl+hLk~Fh
zpy1C<{%LJ^Qh*P%;ZFkpKU*5`M>it=+B!c5kh|ja_!WkMk5!f%NGwXmBid-#je}b_
z_pi|W1H^r%{GSa4SiHX;0;m9_GoQTc0t(9kt<Jc^6}>NgEw~UB!0k~inl9t_=YvZw
zPXPTRz4=jqD^=h&7xeHt*T2lf2NUK0^!v}+^?!=|0kZ#g@_-Zn(f5DZ^&he60nmTh
z^?(!q(f5DZ^&he60nmTh^?(!q(f5DZ^&he60nk6OtKp*y;UFFu)HMmNd)J+9Y>^?z
zFs)uAmMKM)f&vPqu~dnN!iCK#lwgRA>Lr3Qi`7<Ex`$FoX+SBV&?ywm5sRn{Vq^AR
zx=e(WfzttkZ2nRD?FhK44${GXQf*n;qw@bK*TNBrlHjhp7C<wd;6^vdj|TZbl8j`_
ze+Bv0c7HZsSTCE80s=w4eKzmO=bz2N<nzyE^Uh$%31G4`N1Q=tHoq0*-$=w=AYY^e
z`7sIHl?3@iAm2R@58@#IJIJ>Vy0939Qm=8goOC(7AWsAN#!;gl0r`OxO8v&3$MW{a
z@}w&T=%i4Pa3l?)Jug|-+tH^CTxaK%frI`57fJThf?N7n)Cum;2P0T0O`+t+oNYyE
zndNO6U@`{;tek#~{($`t(?3=|g!<=9R<?PFYF$od1#NddDtpxSQCWB{g~Gc9c=OJq
zGW&ZJ%9__Gl=jCTm39AsLTNUdLRr1{vGKIc>dRx-`*D|}f0m&S)4v6H2>JJ!$NKA^
z?e8J)%7pn_^<`P6I>G^P5xOh}ek8$F*6+_s{J(Da7+H_e@hsSka)CG-bY%ishKD>r
z-1oOi%J9(NN%()+>@ga$*}PtZOsjeKD6NO}r!?Q&no{)hE=s*tjVMJ!-Ul@W59Ow8
zV4`Gap1Jd-yxxO6sLy>q@liMdJQl`qPg%AY(HqMg@n|BO2Xi9(QA}w_X+~*H=|Jg3
z=|<^I=|^EuxRe2u=O`kIoT8x^DOSo*ii6^%1SwHUlJXMe70MXOTa<SwQz<hjb13sD
ziz&+~D=BLz8z^5=zNYM??4ul}oS>YcT%cT~{7U(a@<%~IL1{sgf>s3`3OX0`D4-TF
z3U~#B3q%FV0)2tC;Kc$@L8u^J@KV961#cBhDwt95VZq{p6$R@GzAV^Lu)pBDf*%Tg
zEVxl{r?9B7abfGij)gr7p+a6ET&O5C7QRsEDI^Lrg`*3{7rt9KxA5b_)rDUc?kqf9
zc&6}b;jJP{QKO=^MO}*e74eFYB2AI4$XygI8d)^9Xj;+SqUA;Fi?$aXEIL#4Q&CO5
zl6tM`mDTH4Z$Le1J+z*q9#L;(y|?SlsJE!z>Uvx29jy05y<h9yD{fN!Ofj{%yjWIj
zDfSd6i^mjCD_&5%s(4%R;o=L$zm=4fv@PjX!Yz@OSWA2*!%N;SnO#y<@<qx1l5-`u
zN=r*Slu}Cvm1;_{QnGYR>3gLgmu@aSP<o;C_xg?McdpN<FRpK^AF4m9{<Qi_>Tjxl
zp#G)$e>7;`phtu92AT%W2I&S98q903zQNuG7aIK0utmdO4F@$eHuN=orQy2`mp9zj
z@MOd4MvWSEZNzV+YlJs?xzW3gK56t#qaPa8G;Z0rZ)2qKkjC-G;~Ot*{6*vM8do=I
z(xhjT=bG4>kWI!nS={8SCa0SG-n3QI{!L{~olPs7&TP86>4B!dG;7eTM>DwD&}Nxt
z)0%zSY;Uulnm1_PvpLcnYd*61%;xKwA8mfCMXMI{7Md2p7H_v$(qc!8%PmV=_G~F?
z>1;W=<wq^QYI&|zQLApPkXDXXueO@kYFn!dtxH<>YAtK+YyDR1Wv%zL{<TfZHmo+L
zHmNo<+iYxerfpH%o^54q18pa?UD@_X+dJ(#wu9Tb+P&FsdAkGcZnuBB{d4VI?Z>vS
zYJa%>oepIkL>>GcCU#iU;rplRJ=OOq-BZb@-hXP_Q$Ic3`so2rJDwi<^vb7?KU4Tj
z-)HpC41eaMXLdbvt7E5*l8%v%?{)mD<4>L1c6zRpuhW!Hn>zj2xmD-Eo$=07I&bcL
zrL0XETox#Mw`^<KuU($$BJC3I@?n=fUH<6Wt1H@dRM)Dm-*;=+jn~cHZA!N--G1%f
zxx1?SOWl`tKi;E34_*&%kLf+O_qg4&S5I@#*L$w%dAV2nUb0@9UQ2tO?A@eyMQ@__
z+}?-!l=R{C@%NeCXJ6mKzU;o<zBBvor4~{-RGd1Sy8qeYXZg>Do}K&b(SD8l!Tl2b
zmiGIhf7|}b{-gS@>HiDV9kM|aq3zH;8k-iN&7+;5x1`JIqv-4DHyC{x4#rHzA!ZY%
zm|4kO%e>B_vRth9S;yF|*edqx>@DoO94;ryS<d;9+nxI&cNX_3uQgB0dz<$SznCxJ
zSMoRTYsxw0WckYSYXkZX@DErr;PSv;16>2>4Lmof>mY2<hl9=xE*t#f;133$spwLH
zRm`n8`&_r@oX;(I?h@Pw_Q4;+zX)i8s9=@gHd2m^K)w(b2_?d@!absvB2+X(bV}S+
z>=7>&|0-cihD*MX7E2Y<iPEF8jxvXAvFuklS3W|%Rnb^sP|Q%ARrXazmFrc7D!FR1
z>ZH1dI;38sp=e~9$(rxAy|je(b6tr}tDC92sHf{+(r-7kG7K{;G5lr}7$+D{puN#J
zy2aGYG{m&nblWU6PcomeK$a1foz|zUxOJ_q)Mm0Ru-$rI^!(K4FAQN188hU_3%y=Q
zy|82GQ$qtoHx6qy3>)_8i^VURUtIEHEvCWdVz=xv`z-sf4#Y9tan<>pbBgmv*C5v<
z*CqEL_ayga&mhla&yU^;?=<gEID*f>uluCFIlkL|jenuPE?^E+1?vZ246X~c4EaM_
z!=1vZ@csxjGA43{C?}>8zeW|&1+jwI3$e9iYv4@niuaDc7XKkJI59Kvd(xErG}R&%
zO6^MbO~0AGlo4m<53e`eF?`EQU0!<Sr5{GXBR(8iFw#D9OJ%pp(Uliomb|>^m4>hQ
zU)ep1HfqwSo3EN*T|c_x=$A*Idrk7%(lJfPM8_O`eZcE;-l+Em{>Gj+nQy-L=Do4b
zvERN$du!TTwQoD#{&pOF+`HrI#=FPwnZTLw{yQb_gx)zav0~z)Ni8O2CY_tCn!I{S
z*_5$UZcZIGb;mT;v^mooPLEGN^RDvUweNL*@16H*XLx5Eo++4FHS3vKZ_c_s+d2Ed
z`_H|<d``zXZ_T;$0sg_U4<#S2p4)Tow2w+YN_=#2o@w6p`Mmjy7d*A#tp#@%h8CV#
zq+hgkF?aFeCC@CGu(WV#eCdxL5BYfCGU2kd%lj?=u&Qm<Tc1!qiGOls#fvMBtW>Pr
z@+t4r<)8KVY}Tq)tH!P_Se;(|>l*KxGi$AD_pg($+p>PZ`c<Fz|9s(wE*oZUY_oB~
zriPnFZ>rmz*?jAZ$QM6->G|^9mSI~?d}aCS&{o~nJ=+x9zWG}G^;g^B?VG<D^v#AH
z<vZ4W%lmfCPR`ENyV$!{?Pl*@wTH83^<M7Ywfp${*6$y<f75}A1798#9{l={?9k4`
z>cjhwphu1!9dh){G3T+%-vz$Aemrsfj}tGSEIv8*`{v(IJN3+|xu^S_t~$dxv+)Pv
z54+A9&Yn2uICu4Y?0oHo(HEOsoOY@6rNx&SmpA+<{&C-x=dWD28oBz%Ph);=@$;--
zdjIm-wdbzw`qlF5h3nDlbvNFw?ohq(ChO*wTiRQvehdEg$L+Ddcldo#4X<YV9rK+_
ze<W*5Yu~-w=kEG@ihC#Pf^~IuzfhXye&~kc`~vs_1$8GVqbY?CQxV7&)dQbJ^@{6d
zQ*lXgE|rv&mewyVEoso8VZ#Ou8b6r|ii(Pg>lHUBDQVE8QG-TJnl%NfS+g9_{jbLP
zrzhJze5(6`(xN^^UeK$kparF{ML|)Eg1W5*z@ep-0#_m%{p3@<;-bQm0!nH91`Qh(
z0J4i7uLT8w>XKY#b4no~s<2*RNof)ID2x>p7PY8HDelse4(m!<VO@z=S~DhoSlVWV
zpxc-F{Xa0<+JEgH9d*42!ZJ*nyV4Qc!WQ<L{L!bJc19=p)q%71d%oJBVfzlFr-;P!
zJ~JIedp~vKT=mvN=Wn{=uT7o5YTMxpx7>*_(-y4$`pCuKzzJGU^7ZKp*K9v}={8Sl
z@uuE*chTB!j$Qt}4_{`*({H}Fc-@Zgeyr(RF1PtIV`nT`|LyTBcfi2GFOC%z14fqA
zt6x+U1mkH@R4+?$u?|pNLcB7O@!^VJTYtI#hi-y4dVBObvpWkJ)pnAh$6UwCEv2yo
z?6bmG+ZiVV?zHdFlOuYnmy<-fH$M9GtKNMMo~s@`W!`5`k}NhoBKmJ>{s)rlj#C;J
z<;ZM7si6E(GxqZ~iM~If$Cj=z+m~;9;U6BWpxPs~vu`xJ#GO8K=c)AmNi|)YlFMv;
zB_lBE%Z;ZU+&6VVq4Drjug>_S{nG0d_oAiu2<G0?6K}UG-ypr`uicDS?B8%0?{FdT
z@fUrEZKYmbB`sK@Tm4bVKta_^vWk&@3yx1mj;`wXl<@oRE7s5W{x#`<_U~!5P5a*Q
z-?(-!X0E>UjhVAEeQ`|nkqIpdNA%j|xEi|p!Z#a^pJ$i1>ej3cZ|SXtH;V2>{oib!
zv;L+1IDE4i58t0TzIXC3=bQDFXcr7mKc^cw<lVwAN(^fGD%+Af6>Y5DCw}$Jq@7*2
zwU!UH4tWhl+jO7&3B_LZ#p#-z!)j`q{5G-5cMq<k%zEMA$NkS7xG-(6w)Xhpvv&>|
zc8y-jndp>`o7ny(re^oB&xWKv?^`>zdSUIwUbPo0E*9@PI@nu#uX^JbvnKa=ZG086
zz}nJ(V5i1EoLJG#d)++4c8|ChI<~jt;prC_&U*HjxmELHCF?4mdFz`tEZ3|STIsf-
zb(HfJHNQ8jy$D@6|Ep}?$d_uSTgUV){%ON*MtW4Aw)3Vqmh{;9vS0JslvnnRZH(SY
z))d!KDi+=fJxfA0-_%id9$WUoJ8xX5uA@wTbz-NR=5+^_cRV&Wbgl8Ndlron{iCub
zTr;PpvS#yzK^1#DPa7FHSbO0@P+K|o%JG+Gy&Ng<E*LjzN8_c##`4>|kYT)yzW;CZ
z<4@XCRZ)HIV#S53cYf+#N16Ki#f?wHpKW||-!Y#-a%ITYuijc>dtWcF&+4K*TeY=w
z&5rgJm)hUzC?7@}DN2vIaq72=zAHmbFK{k)?6+=Ozkc(kH){2z=Jx5rnd6_KEa+P!
zY2)q$SAV^^CVZ($>7G;8^_=xzze28Q9!kk)cORoIpSMU(`<1u!PWSzl<(naP#m**A
zSNrNHyG*vyDe>W~b^e_-lPbgxbkd>QxL)M_eNp#MeuuXWd$*!`#+>Rowvtb4detmG
zAO5yqFa76htDC+xcrtP6`Hyb(KD%l?waLK~&+prH@{C72>>$;)@4&lDYu<Y}ac)kp
zJ=d=GLg$N~+Oyj)HkmN1_RYnL!4n4V?=@?~WY)*5yY3y&pBv-+sNOq$22c4SI&On;
z<;tq7S0zBjUcS+Ea~*|O!>v94{<uM3e^fJ~bn>+g7Z_W{9`*YdoKO(yfaAjUqpy81
zspn8*%RX(cjotXwznQ!8{<U9zaIC?&-_~I6e!F`yPj2g6WYevqEx5;9+G$<w?%LHW
z<KwHV8|^fWKR0RK@l~^&<EIwPZ@;v|x-31lZ+-T*47${A(s$OiuQXlfU>ZjyEeq@=
z-{VSj!uNHQJ%-9X1HXH3T6p*D;$4k)+->lFM8tooXY0t<v!^ZiJig8H^U>Oim9^&v
zEo^447L?wiZr^_}ZoOlun(sbz^>iD@^vdelr>n;uIqPUP<5r{Ur*&<c^!&X;F^=Av
z`?h{&t&w&6M4Ojp*VNQe_;r*;ej0arMYjW&1S@Bh`v#MnUNhZ#tI4@fUwi9~m!IzN
z)UshKTfQ>3Y7eSm)|?m;UAiRw>8=?czjhwGI&A34Z?-2K*S|Uc`;4<yk^PmIT2~BQ
z^{(kmjf(Zg5VYw>>2RYtvjpOamEZN=raL~N_H4ULQ{V3Ced%-05~AK`oEfjJp_P8N
zx6Mew!ZxX1@7B)0SlN5%z46uRS1LakI?*`lPU-0NbreQh3hlOD+;{4ryR}CW$DPyH
zwEXe6NwR+Lzkgr=(t0w{av(8zoAvi)QV>A>ICj&-8>O{B)KM<ByEycHcXg8ugVi@S
zEiRDtJlxH*ZtW;n*`0(BT?Fh-f?*6%$8Lr;U+TO0QtN@YfPQSd#Q*GPeTQ3{yZBaq
zn&$2$wwO*`Vem7pu0OkCh`4Wu2B&5nSwfh<O+B#ao9E4`+1io2D^po}hg-cxMarj@
z&zW4wtQxrRnN_V1KCk%v?J1>mR~?Fda(&tGvsDLY&Z+sfbItahlA9N5Pu!^fWGFi3
z+|kx`6k88$nD1yCN3x>&y*bt6+P+Z*U3~5Hl7Y{?u0Gt8eRb{F{cqIFLdRznBT>1B
zbK=ga2{W%%-K|)08E<#)P#wj5WK6~HKKsXsOZP@L0RyR{6!m_iX4*~KVlR29V8n!_
z13kZNvwVNFwzl%dP*!+JTCrz3W6az)`;ouDGoh2<?%vw|eYgLz`O<~K&w+#ZHN?e6
zqZ|Foy507ZNvG;j74Wqo=(v-0l)bxtYGJ-oIXT&CMMe9WFGyeRv@OuR>ie-Pu5PZW
z8MF4@nnRVne+t%A5?8Ob+`av7h%+&Ir_b0GC+5`DR9r+>-qqZ=y;Iw}bNd;J9d8ca
z<{CdI(f4c6>A8yc4Ig#yH1x&`vu#VCO>loYlNwpAbyWZ((+P-w&dZg11SeVJnp}Lv
z-qE|b<lT8)=J(z7@ua72pPV>$#piQs_GoH$yt$?RhP%bnpSnT2e8+lWSZC7__4~(`
zcKjU7wJ*T`qz$baY-U|Mc$8f!IP!JF@UYkSuU1$5gagv&)a<CL`KHkt^YrT2u0=+o
z@adzY4a?8=FPjs~V&8LmkN?okYZcsqKgQejA3A0IZ*#8>>(nzk;OxgEuU41tU9-E}
zshZQnHl)pOFB?2R-o-V)_qKWL)+c~0xwmhBSbX89+xGHzw;o{4@*kP?=6u39kl4Ze
zSSLR4&BZsLD%_>{<mt1Ew|25sciiV*wU14miio|-+>vIR+whhQ57)SACMUj~wt4y~
z*E;tzQ|$-B&rCd9xHYlOFtc(;=RJ+4nkL-Xxob)L>xD0H&UPC{bnR1gx{W4SN4fNL
z9c5sD#j};uZIh<98N3v3ey8!IMHS+{vrhmO>V0p-$&9DA(Pc@mKUM}6s-fL#=ik{+
z+_qQ2H^Q|)gfEJhjr_G{%k#tf{J3Ite8YHe*g5(3iZ{Di&W?EV?N6z#)bfJPZ3I>O
zPnV6f)-Z4ElunS|7&ETTD59`h`(meGetg}$cGh^c?9?15sVzCSbP%xM^G0Gfj!v&V
zG5@4@lWYIa?2Y};kNoZ1U)U3=-%a~`Y13WvS{!-?y?9SwJF0f^4QIx4d2^o=L%zR!
z^mh69_ZFXiB^GOav>rKe%vm;S`N}+@`RUNd=<uF>w-5a8LNjX3wVijSPkrga2bP+d
zJ7;VV_mfBtVb(JfhqwH=NeA_nEpIP)P1<$s(i@Gcx8EB<o$oW7ZKuySn{{yKMf-uu
zQ3LlJ*zFxX<NE4DU1dE*MK7-0-lfV)**0v(y$n#CF5rJ*?WSq2!*!INtG>4#?<zes
z@bi{e-&xu0w&#!W`1fsR*81i({lH#DIrP%2mzH%JCNbYIZJfV*$H4;|4u(Dm>E7vx
zclY_`{W8G*@zR~0Cw87aGPZW?iElSg*%UrHR69?Ge>8N1e3_iN`^SuT!AED^MI+ZP
zJ-xj0QvEs#XW*z!LqW%UR`}Oxlb4rH+N&CK_UWbXH6Efsm?t(&{6hYTuWdj1*uo}l
zMt)wohcVK70?g7q14dR>?iu8_bN7D+Je4qWXu9W)%iR>5gQ1ppz7!>G-}FGAzSaj=
zN_cbalG=R-b~e%6IK8`^5H~z~RkCr++oI<N{?NL+{o}#aszc<(bBioPj;(NO*R&WY
znRNF<NfB|n4g8q9Vk=o=UbaiG=s0*+sGG0$+lnjOtef6T*9>jg#If^z{KZA}#(9=m
zn+7#+ywY<@YV0q?&u`gy-TWcGrjAnmw)Dd|1dX(>xmN6a{w~Ax!>WWO`1PsI<NMwm
z#54ld_e+*s-T8ie^dQO8M;FOM&F=I&*zrzs+t63%XH8+r#0KwgxpGSQOwFl#bE47;
z*R;}!4S>bj*i2PX{px|zwwr(0zjNHkb+>L^wHdA&7j7Ip3%#6i&8a=xeO!l?ou@_1
z=g4`>yC3ZE?!IO#2DJP>v+wTjOZTyM7v1RU67K&@(?vdB(bf7q9%#@}+M-IQx^5bd
zUpW1&u-<RwZKpS%nwZhZ(r(i1dk2=EYok6lo~N^qWshaGmhLOo4tn`Sn_>Up?`QqF
zO<^#=75kpOwRuM~*52sSTjL59;Fo=-F8)<Hd7ess@MjTq`S#5t-(7m+)o;o-?isi2
z3YW9z+oOYrRCgTxAso`3y1Zufp87)%O<Hxi{I_kx4sPnZpE9ccq5;hb6MMw@&Aogt
zE}ws4(9>EI(xDxv?Tt~dFRAq=e%odl2b|ww=a(<&eD(B{_tmq{Q&`)DE3DoHA2ev&
z{Hx2nrOh@0x^&<_aY)APJmt=_FU@)H&Y~GdGvJ5Gq?frr9&CfH*<Af9Sf<(y-*&#w
z&5OUUW3R4S{LAwjd#rn>Vd8^rmiX=3KMvk#_*3VldoPabzj*7@L2=Iym!+ExJs$ic
zHQfHOppJ5H#F`p&uZsQe53hb<HvZ#X@3E5yzuwUAgJr|XJA|q8@H)!i-3R9Nx*5Jy
z{QcbO9-=*OOfP*MGj4pnvhu53|36o7zOFc#yf|$l={+&z$LbT;{^)jQT4-bKpzEfN
zBS7=BwLiYhJXW)C$I&ZG>Lc+Hjo<dKr5u}jN6{F4V;E4L!pc2D>-4sFhQ7JvdRd=U
zhr1310m=vr`*Giqcb7KXm{mUTKYfM0^Nl^vkG_a@Uw!E2x>pUqEt>9ndC<Cov7hWb
zUGw#Xy?+$<tGLkay?F(V>L?#sc0|1!6EE|tUcLebQCTCaSvtS8e9ndGGg&{hFG1pS
zPa|9DAM~2lb!x8wpu)TU>uDBWMs;zPZq(ZKqom%g#-iQg^Y%s3{>y`wjjOh<-t*O4
zEVXvx8TrJqE1%a?q_2E&Wc!{Ow<a#~eYxm+$@-fWjbB|DKqns`nHm}Sj<{LLeCeCo
z=ij+?tHUqc+qY(488rUb(w?7!Va%@mZAHT_wOc+PNqObgTYl>J1)ty9H)!N6bg~Vw
z;>COE=U?u5^e1RS^9>6R247ipx9a=u_JzM}v#P3UFO^<AS#N|xd}G^{w#Dm?ztD~N
zrS<&duMBJ1wC|2(6_*BVI5lhV=JTpf=L^$!+O=zLOcYIDs@dKqvt}Y#hBzCReLV2$
zrEvevQ!rV-A69Qzz0O}pK`%~zYsDLI3*noCdxezvC-vT;n~(pNT%8#=f8)J5@isqC
zJ@4-~|9sPxqVA!-ACI3ms{Unh-!~t|0mqhpSx4bkUh28ITSd2SuG$^7UmYZ0&>Xzg
z<;KazN7v6Ao6tYIY?^GU<);o)&9%$M_pQAHV#mAHHx8<1cwc(0Msu+<=h*<-|JGY4
zf=_WjESmJjWbMdaQGVY!%?I?oPOYk`?L6mww;$Ru0XlKw9iE7@FwT5cb>(X77q8t@
z_q_@A0iMtgpQS@BI}f}kJ!P@A?X=*-5B)8^|EAr>excy9SGwI}+N-cyV4-FmuAN!?
zQ|(7}l&1I0>9NnxPVPFYUB2$Zi9eE`ZeH=(DspYCD_gftx?&ml#Vmf)=M8OA73Zra
zhc<k-q~@3FT|5VCXV+9-+#NGc2{addbn?rRTOB<!27NyEx8;5BO}ze3n+i*i;TMKc
zuZY)An09mHbBAZF;Y1XlkND{NhbM1uvjnf5fBpQPdG$Bd&c6}g(OWCM)A4oM!j+pp
z+A;Fgv7h`=N2%EsuGv0LGUrG|pU-$T2d5u)Ogj40&4q_gl@C9-yzf@vGS}aH=_yF&
zs@*WWLx&pHwBxH6tvWtSz&=>kxkvk@O~CA#Goo@&#S3+m$R_8(!AJkty_h;;`JB^V
zLEruS`_VSo(bLsG@BJ1pxZB}UCFgjj)`obG*S2-qsA(ka+vmjfZI%&rl-fTkMDJ|A
zym$AB6Ex|~ZAZp+++EDAYP6PWIJofO*iGEpyMVvC+AnM0oPKl4>FPD<uD1FwopTh4
zC-PZ0uf2Go%Np>Lpw@{?DAB{(rF-fqU^Wi~u|x&6z39TY$c}l_AZU8{(8*&f22ASK
zo#nhuUpe{Y*{WrqX=?U%zd3c_j+(@X$lw9XICYdW4)wm>ZOoOwe{y#5;qA4TCQLih
zle2mH@hJyK?uwrJs^y%S&qiAS=c%yvYe3Q$wcl1R`}p|K(eKo>lC5q0>B#XD=WQax
z-*E%2tM;&wYpNB-&EL$ap4R!jQ`Tx>X79VJWE;-S8}j_~-(8u}@#L)5+d~`L`lE9v
zs=lA``g^@*s~Rr3ORrsDd!cqy?Y<R?In=8wp8lcp;^YS}OdWIY)Tm{9T7LaOC8O=I
z8!OD`PG>{H+6lJxhgVJb_FBz$mH+0@arU8mcXgg&di`$SE3O`{w`7y=9&UN_&X(Fh
z@{J|0CDGpx4<DXhJoJVC*NB-0HFZUCJmb>EVkvaeI=G}*DHf3!febbR+JORM+-is|
zvV~d*DMqwu5(^2XI4G2`M3#b1q-?UpSeC?MFjFK%5+F*zh%B)W3>Zc*0nEJQ<)wF=
z=}f0z?1y{5ow@g(d;VwnrFj%i=&jaaxmBj$9nenp&28`BCW!VYY%3*4DUZBjSl6pb
z8gYne%GL)(M{A<iNc9Bw8-j*wIRxr5kYtdIS$`K~UL7)z%<1wqv-~~0I2g)0&)o{$
zB-flU#T4($=^&I})XMBjm+T}ThwxvTehnj|eEKBE^tSkPjllTWNX&<(D6X4IZbA1<
zqifzv1$e=Y2g>OX7OY@KxgJgVu>=l#uzKf$IQiUq>E$}1napI_EA_mRj$|w^98HBu
zmihj!)~b{eML?c0v9M|9id5)T*v4wMwGBa^6Kj6iEix+OzszrV{BB+|Tpc@{dAnxp
zPE9ML;>@FG<wqS&DOY>GsJ+<~40w-iKU@tSKCk~?;Ya^k^fa&X&B{)DaEPZd!O!GZ
z>_{`eZyl!V%EBl=*a_qVpmqF?0*8a)y8+f&6zV&t+8%af9ai&_3%5#_@t&m*Il~a;
z(&V1C0ozpte&e7uhQ<GM6V0T9y1>nhW4h(^Ej`ne(G^BzYe%==FtG|U8rg127s5xr
z3k!czjvp89&sRur8dtcFABe{Cwen6atcp$jSdOUld|zZViT$mvv5o~qF*c=6&q}i;
z(W>V5ZOROYkL&lj<M`iy6L#NyE&UN&neZ0ML3#Y41F5)^It9?%?3T{kJ2)oSDoG+a
z?a=BL6}>(x!`*GfUkzP7LX+Gx3~2YF46gPBT@kR-eeTG}G|l|T)D!8x>b#A3Az9$2
z2n*lPGDwJted_AT)oUN;(0S1oX`_*0OcZ}0I=$q}THNtdPat;kN;c^V#SThvkO(ZU
z6XPRTzz^UCBBUdsyMl@2jnOh@%y?wf@zz4qvC>0oWmLk2{hra(av%FVmiyss<BUQG
z0G|`H1DS*qW=WBT!%HDo1M6?J%+<*}Sem2P<Wl@_<f*XYb=eafX)h$1zKfm>MKh>N
zd1wZIP!=DDs4Nrnwesh<ioGlJjs3Nz-&b*hy9a|>%CM>NrQXkXZa*NYewDKUeT2Qh
zHUEGYL2vKcbT=8Mz^rD*OT2d@+jxma><SdG-=Mv&?OD4u_l3*J`{o!DP4es=^o$72
zYMH^qeP6ZmcV$|+>2Lw8^*Gx;ntL#y+ocQ3($R|F;6~Ec_httY!-}ApmgU9QP^|y(
z=_C~RPO!#cDRN>sp-(>ndx$1W!vE_e#WJZt>ex8jQ0O;K3<#Rd__h#(>_vl*ZB`>P
z_v)OO>1xLqu#T`RUlv(I41Kqn*{_1uq8QR5Wm9LIn9Dc^#6*6@SqfHgW?+#o&!Cz+
zffIl`gzJ`^7_W%&3MXcp6{($!u*MrS%aBAwpw3Z|CkCYcGd9}Jv_AmE^ERkaV76KG
uMq*LBbvj=mX8JlYYSM3fmA?b}84WU+hwQ5zGvEY;a2$J>K&45XV}Aqq_kjZd

literal 0
HcmV?d00001

diff --git a/read_LSstack_fast_float_mex64.mexw64 b/read_LSstack_fast_float_mex64.mexw64
new file mode 100644
index 0000000000000000000000000000000000000000..d5df9b022a2336f90c02a4f9df83f539ef484984
GIT binary patch
literal 7680
zcmeHMdvH|M89%$ph9nR+D626VS;EpXrG&szO-zDWNbsz^!2p3qM3-fA7gqL>-n~o1
zR2`g<;LdSvq>it#RQj~`5A2kwK<l_M)PyEz%43YhDJf9hG@y~TLIJt`opbMoDAl3u
zOlLZB!a3h}zUMjL`Of1mbt^ka1|cL9RnrLB1;|UaXg|Lh1bgC?Jrl{!oc&XGnWX(w
z8`cG)_GTrrR`G`HK5sZ2QSEDFyAliA17Um3UCZpDNRwQWmzV1>OdpQ?{P5!wf7_m}
z7hCRb|1SC$TO#ex06x88Mf;P2zqkFDf*x+)D&T$X9RfBAxHRBf$8CIi#Pzj=G__}u
zcb4C{G|d|!#rE;$Tte`V<m=RW+bF)t#7UfPCd9%Ulr$vG=pb0zGELxAL)0)}fXn+j
zW%ZW`@j<nv7~|mv>j~M(9aR&O1Q;!=37Ih3`8N<dA*EvA|5~R+m0MJ>7UZDeW94(_
zi$12<30YmDG<j7oAzfKufIzZQC!&ri)fjb&E=C$Ty%d#)={nxIRH`wxl4e~n$;luH
zxdPs~RI0f>C93fFh?pQYLay+V49BX4oE(fmP{e38h}`c@sAEbsra7j275$tD(D*7x
zHjM`z`82LN3TXTRhuzgcTTj~EEXU3MA+cjA2l9$CD`f+zYk}@TCDt7SS{L>8rp}`S
zW)172te>(}>VzQNm7O}wk%a9uvW>F+l#OV%kE|T;q3o<?I|kf&peC`v%z89iALsZc
zO1?#zl66NXea786Xr}B<_nEWstJ!`e_$x9YL)mG`*H!D>XWem-T6cWtw(c0IO}NIT
z?h~GS-0@6DisvPq$I{N>a3f4e?Dai7>tR{q<ACM+xn<U?*<1oy<tU{r=qRI1byQRK
zfP>Q3y>{2CAqUj7=_etiP5+~4`$hYjXm^VCC!*~@TfL0Oa5q7$G+P7&F>Sc(#b%;$
zOW(}`yuv}-k(ed1&@?=awoW6dH!wb$$kPB)FY5SvIH6P9fIE|XF!yxz^-;E$#v8G6
z>#%Sw1yX#q-L>4c!u1XA)~VI8L5U5u9k8~QVFb)F3x3*?*2kvt@$}TN58{?WkYvg-
zr}N!QS!4EaKCS3d3wNjBO|xB#aXOPD)cT8}d`QNQ4`;bv;VaK9XNw4FJ!@92yQiR^
zT5BO>XggSx<^3Hj&2}`8M|zOL<K`!Y^}IuZ@j4iF?1IEbkSVnFRJOa~q_UEhKS<OM
zH5)%?Xn8GB|4dmO%sdTgu|)52BwkCjgK)7`Sb9M4@*YG-hy>A>D$d6&QpJe+V~L%Y
z*zw_~+^kQtZ3j!U{ZQDjNL(RqsRM{t%|=_>cFNjzoDXtK)tAv9dJDKSImENK^_<Dt
z)(vDdPV9rrBQ=Dx$cWUtIEfG`ixH?d(QQkr6ZAOUjs+Z^kE9&nLrLUy0(4t`3K<^O
zR3X@H$=GHHd|wyn07!j|)#$c73@LunwVjIPQr1nkSiTJkEw6eMK%EL}tw2(bK^P7v
z)v>*MxVW3`*K7fBG|No1ySO9DR`J9RBC(O(^%84xG>R1WxbAb^i`2T5Tg@W3#BosS
zW+UvJ#QyYaZX@29irYGGgtG&&YszhJ0*mE3kAOV%2H1$3YH2-Zj@?R`<s8C;Oi;76
zel`%VaoF3ASU0}_DeOh|7FH-#`mv@NEF?@~EUO3;(pjyVDfTtQXAaby2e_IEywT3d
z3z{aeq6D1PwH;B%rx`BoV4H@)Xc@6vM;*Qj;WGm_?wi)j-i}KSyLuDGxzreEep)O&
z-r+3FY?Z^_ZRy2@q@nP$P6Sakf{6H)x>kwU>F25H$QL`&&CaqvQ1&4r60@YDSQCkL
zGo<kw+yYJ-6(<e0v}b6`MNQ+ePvphnW#|G#^)rOBw~_VFao4tX+}ac|gw9zlHeMWC
zx#N*m5AEjnbZP)v9gq$S>BdoMme(LCiwp|Z9|vAF@39^!QG<j?G&ssqTcMBbM499$
z(`<GjZJ7#HPN~*x6LrcsP;`g)2!$^FRKn3SBXtMjLlNBMnE|Z_^ch`!;*JsFpoyaM
z+(e@4ZvewM$}@Osn6-Y|+LcW!jw+}0xbTp4eM0Y`uGe=3dI;lap{#Rv-AY3I<=JY6
zGl?sUR?>_fXY%24{xpnD*fSne@Ru(sus)IeYgeD-b5-JMk=Ow#qtj)5xq5P~_2tH_
zI};Kqs6}Gk%$7`$Dh?>Q%$1$-npDxRSY28tZS9;<ThXf=;~rQiPXwdzm`JQsO5|-#
zX_}kuOWltEVa~;MGdta^pHm*;Ap^<EsvVuFDQEgy2Tj9fan>Vx+Vp;`tv3CDXitdt
zUD2k{a{D~F@@Z?;clg^bmZjP50z=>Hqwz1$<IgPD;_*dWn;fNtzB=~@M`t}V6a9*A
zbpk!pkAv~T^;nk;G$A=kagL7zkLRgodk1I@8*n?1Xtsed_#rL=>9RDl7X%49Ws>s<
z-G!N<fHG}>?s6LLq>7`GwPrwy7q!Ei`^-ULn}O+@N-RBQeT|soWYJzH+5)uI%lQRI
z*<UF8J&o7Rpv=Uw<8_YIDeRvPo`>QHONYO2RKDAi2N!%ahR`TS&*aC!`8<6%a_)P$
z40m19`a~zgpJ4|v|35BY!FLHSlMQ%ry54`^fQJmY%YZK#@OcAv7*N+D<Ozq4Z8PAD
z2Ha`DM-4b&z!L^MZNPH|%r)#57_iiU3k>KtV6y=qGGMy_e`nZ9k4ybjf3?wHVZb{L
zSZ2Vl8Zh60X*-_<8z+%X{ANrSWlb8pGH5EyC537ZDQwCng$?6K;k+zTIQP1wDa%Z<
z=9)`QMH5KT-tnYJ%_T(*IizS_HYu8W!^ri6w*DJz*~A9Attp4tK({TOl-!AA$H7K!
zpG?SfeA~W+ZO>sj`ZECsjc+iMHH%oA#t|$0S?7*#HvH93GW>Eod|n5+F23*1rS)<%
zNiKA9=VmqM;9EFz0lwV~*|mVrqS|lJp~>Kzgif~58JKn{F83qGXYGXajG=7;Ed^TX
z7}{?@s|Rhc_>yi7&s)$wt6a@*Ju+9C-{y=^%R)u=s`BDkNLB*AI<Kk(T0FOwmXd|w
z-65+r{0D+o9bX1!c&*^>H$&4K6kq3Sb?#!^tU6XlNHDM_)KbzE43Z@zB)2RNLvp>K
z$~u{NEq-`zB9g`L3r3<cMxNmbMrC<@;Ffu0qS@mQ1m$pq+->nkn&mL5Tef21(mAE2
z!YCP+;Ry-%9-pF;cC#lC4ydZEgvgdlxJM>F?gh3_n>=#Zho7UGJmDBdwaw)5hN5dd
za!UX|l-z9ccour1&9X1x5BNOmyy2#xtPnDV)0>+cea-5kKoI8f8<5B23(DTGN8Q*g
z!>d2y3423w6e##%Z;+6CGCeVUz7*1kK_o2Pd}Q{BD10x&V^H2FwPH~95ur_bNsY`0
zneHbumjz|HnH(cEfv9&)P;OYqV<rXGC|+eFkAIZ>EMrMbRyNklia(-+ykVcbFcRZq
zA<vqzehmTNdO;wh*M$DUn4-vGRX5~r!so^qYP}Ni$<Zjto4D?>jZsw&H3UMkD~kEc
z93x+^UAnkda?aJ);qslB{_Slm_9{P1()AJaA4Topdl(<;FLOxvbUf0ssOt5t_xQb0
z)#DFFys8H|bIZIEwBJH904$1yeQF>Q<~mo)e;NV)AmSzc6x@HW{eb@vf3jUIUx)}`
z-<3=v<dLa_;0;55h59`3U4WDDR(JvUWWYMy|9gS+-^mlG1b6UrfD+yW1;9D<p*8_;
z0(=g&jLQHXGVs-S6KqGV0X+$L9F@zT0DKDXiXd?QPC1O451hYI_<Mun9NO@n;5dI*
zEHLmd#2Wn%CYdRGm<Ynm>!!5JG~EmWeki<*o&y02USmBWD@|BSU4c}3qft4uCb-ew
z5(<W+RmCwSd|TAFP7ZmavqAx%5{X9q>MUO*belICD%min*p8bj;FqK73R#KbT(2sg
zQ&L(yKQGsAuT+&-RCVKKy+UhFT@RX3*%wm+>PBJ5j#8<}-;BYF+*GdwHsA_dD@R9}
zW7)MWP~{hlByW&|_8@PoioH>Hctd2ptQ6a00hf>8ja9{dFJ39d^DAeM+Ww^SsQob#
zmwB(8{fV(w&K{jK)GKGF<2k?HRqv{)aWB53vC%;On?AAY|Cd6yV-c^GFI)uv33lYz
AmjD0&

literal 0
HcmV?d00001