-
Notifications
You must be signed in to change notification settings - Fork 15
/
test.m
51 lines (42 loc) · 1.7 KB
/
test.m
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
%测试用的文件
% by 郑煜伟 Ewing 2016-04
%% 验证AE梯度计算的正确性
% diff = check_AE(images); % 已经验证,除非有时间,别运行!慢!天的那种
% disp(diff); % diff应该很小
%% 测试 sparse DAE:训练一层 sparse DAE,将重构数据与原数据进行对比 - DAE通过
clc;clear
% 用到 loadMNISTImages,getAEOption,initializeParameters,trainAE函数
[input, labels] = load_MNIST_data( 'dataSet/train-images.idx3-ubyte',...
'dataSet/train-labels.idx1-ubyte', 'min_max_scaler', 1 );
architecture = [784 196 784];
% 设置 AE的预选参数 及 BP的预选参数
preOption_SAE.option_AE.is_sparse = 1;
preOption_SAE.option_AE.is_denoising = 1;
preOption_SAE.option_AE.activation = {'ReLU'};
% 得到SAE的预选参数
option_SAE = get_SAE_option(preOption_SAE);
option_AE = option_SAE.option_AE;
count_AE = 1;
theta = init_parameters(architecture);
[opt_theta, cost] = trainAE(input, theta, architecture, count_AE, option_AE);
% 将训练好的AE所重构出来的图片输出,与原始图片进行对比
option_AE.activation = {'ReLU'; 'ReLU'};
predict = predict_NN( input, architecture, opt_theta, option_AE );
images_predict = reshape(predict, sqrt(size(predict, 1)), sqrt(size(predict, 1)), size(predict, 2));
% 灰度图
figure('NumberTitle', 'off', 'Name', 'MNIST手写字体图片(重构)');
show_images_num = 200;
penal = show_images_num * 2 / 3;
pic_mat_col = ceil(1.5 * sqrt(penal));
pic_mat_row = ceil(show_images_num / pic_mat_col);
for i = 1:show_images_num
subplot(pic_mat_row, pic_mat_col, i, 'align' );
imshow(images_predict(:, :, i));
end
% 热量图 jet
figure('NumberTitle', 'off', 'Name', 'MNIST手写字体图片(重构)-热量图');
for i = 1:show_images_num
subplot(pic_mat_row, pic_mat_col, i, 'align');
imagesc(images_predict(:, :, i));
axis off;
end