cvLabMATLAB/SAMPLE_MAIN_cifar10.asv at main · ComputerVisionLaboratory/cvLabMATLAB · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
 clear
% load('FACE_16x16x1000x10.mat')
% load('TsukubaHandDigitsDataset24x24.mat')
load('tsukubaHand24x24.mat');
whos
numTrainingSamples = [200, 400, 800, 1600, 3200, 6400, 8400, 10000];
numTestSamples = [40, 80, 160, 320, 640, 1280, 1680, 2000];
% numTrainingSamples = [200, 400 ];
% numTestSamples = [40, 80 ];
numTimesToRun = 15;
avgResultMSM = zeros(1, numTimesToRun);
avgTimeMSM = zeros(1, numTimesToRun);

avgResultKMSM = zeros(1, numTimesToRun);
avgTimeKMSM = zeros(1, numTimesToRun);

avgResultRFFMSM = zeros(1, numTimesToRun);
avgTimeRFFMSM = zeros(1, numTimesToRun);

avgResultKmeansKMSM = zeros(1, numTimesToRun);
avgTimeKmeansMSM = zeros(1, numTimesToRun);

avgResultKmeansRFFMSM = zeros(1, numTimesToRun);
avgTimeKmeansRFFMSM = zeros(1, numTimesToRun);

tempTestData = testData;
tempTrainData = trainData;
for i = 1:length(numTrainingSamples)

    fprintf('Number of Training Samples: %d\n', numTrainingSamples(i));
    for j = 1:numTimesToRun
        testData = tempTestData(:, 1:numTestSamples(i), :);
        trainData = tempTrainData(:, 1:numTrainingSamples(i), :);
        trainData = orzNormalize(trainData);
        testData = orzNormalize(testData);
        [nDim, nNum, nClass] = size(trainData);
        testData = orzSlidingData(testData,25,25);
        [nDim,nNum2,nSet2,nClass] = size(testData);
        nSubDim1 = 20;
        nSubDim2 = 5;
        nSigma = 2;
        nk = 20;
        numRandFeats = 500;
        fprintf('Running for the %d time\n', j);
        %% Using TsukubaHandDigitsDataset24x24.mat

        %% MSM
        %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
        tic;
        [V1 D1] = orzBasisVector(trainData,nSubDim1);
        [V2 D2] = orzBasisVector(testData,nSubDim2);
        SIM = orzCanonicalAngles(V1,V2);
        RSLT = OrzEval(SIM(:,:,end,end),orzLabel(size(testData,3),nClass));
        elapsedTime = toc;
        avgTimeMSM(j) = elapsedTime;
        avgResultMSM(j) = RSLT.ER;
        % displayResults('MSM',RSLT, elapsedTime);
        % % % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
        % %% Kernel MSM
        tic;
        [A1 D1 C1 K1] = orzKernelBasisVector(trainData,nSubDim1,nSigma);
        [A2 D2 C2 K2] = orzKernelBasisVector(testData,nSubDim2,nSigma);
        SIM = orzKernelCanonicalAngles(trainData,A1,testData,A2,nSigma);
        RSLT = OrzEval(SIM(:,:,end,end),orzLabel(size(testData,3),nClass));
        elapsedTime = toc;
        avgTimeKMSM(j) = elapsedTime;
        avgResultKMSM(j) = RSLT.ER;
        % displayResults('Kernel MSM',RSLT, elapsedTime);
        % % % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
        % %% RFF KMSM
        tic;
        [A1 D1 C1 K1] = orzKernelBasisVector(trainData,nSubDim1,nSigma, numRandFeats);
        [A2 D2 C2 K2] = orzKernelBasisVector(testData,nSubDim2,nSigma, numRandFeats);
        SIM = orzKernelCanonicalAngles(trainData,A1,testData,A2,nSigma);
        RSLT = OrzEval(SIM(:,:,end,end),orzLabel(size(testData,3),nClass));
        elapsedTime = toc;
        avgTimeRFFMSM(j) = elapsedTime;
        avgResultRFFMSM(j) = RSLT.ER;
        % displayResults('RFFMSM',RSLT, elapsedTime);
        % % % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
        % %% Kmeans and RFF KOMSM
        nSubDim1 = {20};
        nSubDim2 = {5};
        nk = {20};
        nSigma = {nSigma};
        % nSigma = num2cell(linspace(0.1,1,10));
        % numRandFeats = 200;
        % KmeanKOMSM
        tic;
        RSLT = KmeansKOMSMEval(trainData, testData, nSubDim1, nSubDim2, nSigma, nk);
        elapsedTime = toc;
        avgTimeKmeansMSM(j) = elapsedTime;
        avgResultKmeansKMSM(j) = RSLT{1, 1, 1, 1}.ER;
        % displayTimeTaken(elapsedTime);
        % % KmeanKOMSM with RFF
        tic;
        RSLT = KmeansKOMSMEval(trainData, testData, nSubDim1, nSubDim2, nSigma, nk, 'RFF', numRandFeats);
        elapsedTime = toc;
        avgTimeKmeansRFFMSM(j) = elapsedTime;
        avgResultKmeansRFFMSM(j) = RSLT{1, 1, 1, 1}.ER;
        % displayTimeTaken(elapsedTime);
        % fprintf('num random features: %d\n', numRandFeats);
        % displayTimeTaken(elapsedTime);
    end
    fprintf('\n\n======= AVERAGE RESULTS FOR WITH %d SAMPLES ======\n', numTrainingSamples(i));
    displayResultSummary('MSM', mean(avgResultMSM), mean(avgTimeMSM));
    displayResultSummary('KMSM', mean(avgResultKMSM), mean(avgTimeKMSM));
    displayResultSummary('RFFMSM', mean(avgResultRFFMSM), mean(avgTimeRFFMSM));
    displayResultSummary('KmeansKMSM', mean(avgResultKmeansKMSM), mean(avgTimeKmeansMSM));
    displayResultSummary('KmeansRFFMSM', mean(avgResultKmeansRFFMSM), mean(avgTimeKmeansRFFMSM));
end

function displayResultSummary(model, avgErrorRate, elapsedTime)
fprintf('Model: %s\n', model);
fprintf('avg ER: %0.3f%%, ', avgErrorRate*100);
% fprintf(' Samples: %d\n', numSampes);
displayTimeTaken(elapsedTime);
fprintf('\n');
end

% for i = 1:length(nSubDim1)
%     for j = 1:length(nSubDim2)
%         % for k = 1:length(nSigma)
%         %     for l = 1:length(nk)
%         %         for m = 1:length(numRandFeats)
%         nk = nSubDim1(i);
%         disp('------- KmeanKOMSM -------');
%         tic;
%         KmeansKOMSMEval(trainData, testData, nSubDim1(i), nSubDim2(j), nSigma, nk);
%         elapsedTime = toc;
%         displayTimeTaken(elapsedTime);
%         disp('------- KmeanKOMSM with RFF -------');
%         tic;
%         KmeansKOMSMEval(trainData, testData, nSubDim1(i), nSubDim2(j), nSigma, nk, 'RFF', numRandFeats);
%         elapsedTime = toc;
%         fprintf('num random features: %d\n', numRandFeats);
%         displayTimeTaken(elapsedTime);
%         %         end
%         %     end
%         % end
%     end
% end


% % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% KO = CvtKOMSMSVD(trainData,nSubDim1,nSigma);
% V1 = KO.TransformS(trainData,nSubDim1);
% V2 = KO.TransformS(testData,nSubDim2);
% SIM = orzCanonicalAngles(V1,V2);

% RSLT = OrzEval(SIM(:,:,end,end),orzLabel(size(testData,3),nClass));
% displayResults('KOMSM',RSLT);