MATLAB interface updated to include option for choosing neighborhood …

…function

src/MATLAB/MexSomoclu.cpp

@@ -99,13 +99,13 @@ mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) {
     }
     mxFree(gridType_c);
     bool compactSupport = (bool) mxGetPr(prhs[13])[0];
-
+    bool gaussian = (bool) mxGetPr(prhs[14])[0];
 
     //Call train routine
     train(data, data_length, nEpoch, nSomX, nSomY,
           nDimensions, nVectors, radius0, radiusN,
           radiusCooling, scale0, scaleN, scaleCooling,
-          kernelType, mapType, gridType, compactSupport,
+          kernelType, mapType, gridType, compactSupport, gaussian,
           codebook, codebook_size, globalBmus, globalBmus_size,
           uMatrix, uMatrix_size);
 

src/MATLAB/mex_interface_test.m

@@ -12,10 +12,11 @@
 mapType = 'planar';
 gridType = 'rectangular';
 compactSupport = false;
+neighborhood = 'gaussian';
 [sMap, sTrain, globalBmus, uMatrix] = somoclu_train(sMap, D, 'msize', msize, 'radius0', radius0, ...
     'radiusN', radiusN, 'radiusCooling', radiusCooling, ...
     'scale0', scale0, 'scaleN', scaleN, 'scaleCooling', scaleCooling, ...
     'kernelType', kernelType, 'mapType', mapType, ...
     'gridType', gridType, 'compactSupport', compactSupport, ...
-    'nEpoch', nEpoch);
-
+    'neighborhood', neighborhood, 'nEpoch', nEpoch);
+

src/MATLAB/somoclu_train.m

@@ -29,6 +29,7 @@
 %   'scaleCooling' 'alpha_type' (string)  Learning rate cooling strategy: linear or exponential (default: linear)
 %   'gridType' 'lattice'  (string)  Grid type: square or hexagonal (default: square)
 %   'compactSupport'  Compact support for map update (0: false, 1: true, default: 0)
+%   'neighborhood'  Neighborhood function (bubble or gaussian, default: gaussian)
 %   'nEpoch' 'trainlen' (scalar)  Maximum number of epochs
 %   'sTrain','som_train '  = 'train'
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
@@ -81,6 +82,7 @@
 sTrain.alpha_type = 'linear';
 sTrain.kernel_type = 0;
 sTrain.compact_support = false;
+sTrain.neighborhood = 'gaussian';
 sTrain.scale0 = 0.1
 sTrain.scaleN = 0.01
 
@@ -108,6 +110,7 @@
 %         if l>2, radius = varargin{i}; tlen_type = 'samples'; end
       end 
      case {'scaleCooling', 'alpha_type'}, i=i+1; sTrain.alpha_type = varargin{i};
+     case 'neighborhood', i=i+1; sTrain.neighborhood = varargin{i};
      case {'scale0', 'alpha_ini'}, i=i+1; sTrain.scale0 = varargin{i};
      case 'scaleN', i=i+1; sTrain.scaleN = varargin{i};
      case {'sTrain','train','som_train'}, i=i+1; sTrain = varargin{i};
@@ -137,13 +140,18 @@
   end
   i = i+1; 
 end
+if strcmp(sTrain.neighborhood, 'gaussian')
+    sTrain.gaussian = 1;
+else
+    sTrain.gaussian = 0;
+end
 [sMap.codebook, globalBmus, uMatrix] = MexSomoclu(D, sTrain.trainlen, ...
 sTopol.msize(1), sTopol.msize(2), ...
 sTrain.radius_ini, sTrain.radius_fin, ...
 sTrain.radius_cooling,  sTrain.scale0, sTrain.scaleN, ...
 sTrain.alpha_type, ...
 sTrain.kernel_type, sTopol.shape, sTopol.lattice, ...
-sTrain.compact_support, sMap.codebook);
+sTrain.compact_support, sTrain.gaussian, sMap.codebook);
 rowNums=colon(1,size(globalBmus,1))'
 globalBmus = [rowNums globalBmus]
 sTrain = som_set(sTrain,'time',datestr(now,0));