Add Ensembling Code

This commit includes code and six pre-trained models for ensembling.

Discriminative/ensemble.m

@@ -0,0 +1,52 @@
+%% Ensembling Script
+% A Brock, 2016
+%
+% Take the outputs of multiple models' predictions and combine them.
+%% Excavate the Playing Field
+clc
+clear all
+close all
+tic
+class_ids = {'airplane', 'bathtub', 'bed','bench','bookshelf','bottle','bowl','car','chair','cone','cup', 'curtain', 'desk', 'door','dresser','flower_pot','glass_box','guitar','keyboard','lamp','laptop','mantel','monitor','night_stand','person', 'piano', 'plant', 'radio','range_hood','sink','sofa', 'stairs', 'stool', 'table','tent','toilet','tv_stand', 'vase', 'wardrobe', 'xbox'};
+%% Get Results and Models
+
+% Targets
+y = csvread('y.csv');
+% Modify targets to account for the fact that MATLAB starts at 1
+y = y(1:12:end)+1;
+
+% Acquire model results
+names = dir();
+names={names.name};
+names = names(3:end);
+n = 1;
+for i=1:length(names)
+    if strcmp(names{i}(end-3:end),'.csv')&&(~strcmp(names{i},'y.csv'))
+        models{n} = names{i}(1:end-4);
+        n=n+1;
+    end
+end
+
+% Get model accuracies
+x = cell(length(models),1);
+w = zeros(length(models),1);
+% Get all data
+for i = 1:length(models)
+    x{i} = csvread(strcat(models{i},'.csv'));
+    [~,yx] = max(x{i},[],2);
+    w(i) = sum(y==yx)/length(y);
+end
+
+% Sort model accuracies
+[~,order] = sort(w,'descend');
+
+%% Get results!
+
+z = zeros(size(x{1}));
+for i = 1:length(x)
+    z=z+x{i};
+end
+[~,predictions] = max(z,[],2);
+accuracy = sum(y==predictions)/length(y);
+
+fprintf('Accuracy is %8.8f, with %i correct examples out of a total of %i instances.\n',accuracy,int16(accuracy*length(y)),length(y))