SparseDeepBoltzmannMachines/Image_generation.m at main · OPUSLab/SparseDeepBoltzmannMachines · 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
clc;clearvars;close all;
% rng(12345);

% load adjacency and colormap
load('JJ_4264.mat')
colormap1 = readmatrix('colorMap_4264.csv');
required_colors = length(unique(colormap1));

Groups = cell(1,required_colors);
for k = 1:required_colors
    Groups{k} = find(colormap1==k);
end

NM=length(W); % number of total pbits
num_pbits=NM;

%sample image to clamp the labels of it
image_index= 169; % Change it as required


num_sweeps = 10000;  % number of sweeps


beta = [0:0.125:5]; % annealing schedule


% % Loading MNIST dataset
oldpath = addpath(fullfile(matlabroot,'examples','nnet','main'));
filenameImagesTrain = 'train-images-idx3-ubyte.gz';
filenameLabelsTrain = 'train-labels-idx1-ubyte.gz';
filenameImagesTest = 't10k-images-idx3-ubyte.gz';
filenameLabelsTest = 't10k-labels-idx1-ubyte.gz';

XTrain = processImagesMNIST(filenameImagesTrain);
YTrain = processLabelsMNIST(filenameLabelsTrain);
XTest = processImagesMNIST(filenameImagesTest);
YTest = processLabelsMNIST(filenameLabelsTest);

y_label = double(string(YTrain)); 							% converting categorical to numeric values
y_label_test = double(string(YTest));


x=ceil(XTrain(:,:,1,image_index)); % from training dataset
% x=ceil(XTest(:,:,1,image_index)); % uncomment if you want the image from test set

figure(1)
colormap gray
imagesc(x)

bb=x(:)'; % Flatten x1 values and taking as row vector

t=2*bb-1;


load Jout_100.mat
load hout_100.mat

load index_visible.mat
load index_sticker1.mat
load index_sticker2.mat
load index_sticker3.mat
load index_sticker4.mat
load index_sticker5.mat

index_sticker=[index_sticker1; index_sticker2; index_sticker3; index_sticker4; index_sticker5];

J3=Jout;
h3=hout;


% % Clamping h to labels value
label=y_label(image_index);
% label=y_label_test(image_index); % uncomment if you want the image from test set

hclamp= zeros(1,NM);

hclamp(index_sticker)=-1000;
hclamp(index_sticker(:,label+1))=1000;


h1=h3+hclamp;

% save h1.mat h1
J_bipolar = sparse(J3);
h_bipolar = h1;

s = zeros(length(beta)*num_sweeps, num_pbits);
x = zeros(NM,1);
s_temp = 2*rand(NM,1)-1;

tic


for kk= 1:length(beta)


    fprintf('running \x3B2 = %0.6f\n',beta(kk))

    for k=1:num_sweeps

        for ijk = 1:1:required_colors
            x(Groups{ijk}) =  beta(kk)*(J_bipolar(Groups{ijk},:)*s_temp+h_bipolar(Groups{ijk})');
            s_temp(Groups{ijk}) = sign(tanh(x(Groups{ijk}))-2*rand(length(Groups{ijk}),1)+1);
        end
        s ((kk-1)*num_sweeps+k,:)= s_temp';

    end


end

toc

mm3=s;

% energy calculation to find the best energy sweep
%  Energy = zeros(1,length(beta)*num_sweeps);
%  for ii=1:length(beta)*num_sweeps
%      m1=2*s(ii,:)-1;
%      Energy(ii)=  -(m1*Jout/2*m1'+ m1*hout');
%  end

%  [min_Energy, I]= min(Energy);

%  figure(6)
%  plot(Energy)
%  nn=mm3(I,index_visible); % reading from all beta, best energy sweep

nn=mm3(end,index_visible); % reading only at last beta, single sweep

B=reshape(nn,[28,28]); %reshaping to 28 by 28 matrix
figure(2)
colormap gray
f1=imagesc(B)