-
Notifications
You must be signed in to change notification settings - Fork 0
/
autoencoder.m
27 lines (26 loc) · 854 Bytes
/
autoencoder.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
%X = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15];
%X = X';
tic;
X = zeros(16);
for i = 1:16
X(i,i)=1;
end
%layer = reluLayer('Name','my_relu');
%layer = reluLayer();
%layer1 = relu();
%generateFunction(autoenc,'C:\Users\Devanshu\OneDrive\Data Mining\HW4\myfunc');
%bias = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0];
autoenc = trainAutoencoder(X,3,'EncoderTransferFunction','logsig','MaxEpochs',200000,'DecoderTransferFunction','satlin','L2WeightRegularization',0.0005,'SparsityProportion',0.66666666);
Z = encode(autoenc,X);
XReconstructed = predict(autoenc,X);
%use tied weights
mseError = mse(X-XReconstructed);
enc_weights = autoenc.EncoderWeights;
dec_weights = autoenc.DecoderWeights;
%autoenc = trainAutoencoder(X,5);
%Z = encode(autoenc,X);
toc;
%net = network(autoenc);
%options = trainingOptions('sgdm');
%convnet = trainNetwork(Z,layer,options);
%10084