MATLAB: How to get the encoded data from a retrained autoencoder

autoencoderdataDeep Learning Toolboxencodedretraintrain

How to get the encoded data from a retrained autoencoder?
Here is my example:
X = bodyfat_dataset;
autoenc = trainAutoencoder(X);
extract1= encode(autoenc,X)'
%convert to a net work
net = network(autoenc);
Xpred = net(X);
%how do i extract the encoded data.
%tried this but it did not help
extract2 = net.outputs{1}

Best Answer

Here are two options you can use:
Option 1:
1. Run the code to train and retrain an autoencoder:
% Train a first sparse autoencoder with default settings.
autoenc = trainAutoencoder(X1);
% convert this autoencoder into a network:
net = network(autoenc);
% retrain autoencoder:
net = train(net,X2,X2);
2. Then use 'genFunction' to generate a function for the autoencoder network:
genFunction(net);
This will create a new function on the current folder called 'neural_function' that contains the code for the autoencoder 'net'.
The customer could then edit this function so that it outputs the output of layer 1 (a1) (I have attached an example of how the function will look like after the changes).
3. Finally, the encoded data can be obtained by calling 'neural_function' on the new data:
features = neural_function(X1);
Option 2:
https://www.mathworks.com/matlabcentral/answers/300167-re-train-a-pre-trained-autoencoder 
Related Question