train neural network while second trained network is concatenated

Hi

I have one trained network (already trained the network), let call its Net2. I would like to train Net1 (new network) ,while net Net1 is concatenated to net2 (Net1first Net2 second).Net1 output connected to Net2 input. Net2 input is fix and not changing. while only Net1 should change.

The data use to train net1 is the input to Net1 and the output from Net2.

Net2 = trainNetwork(XTrain2,YTrain2,layers,options); % Net2 was trained according to the Xtrain2 and Ytrain2 data

XTrain1 is the input to Net1 , Ytrain1 should be get as the output from Net2.

Could you please suggest a code for doing that?

Answer 1

[Srivardhan Gadila]

I think it may be complicated or may not be possible if you are solving the above problem with layerGraph and trainNetwork based workflow. Instead I would suggest you to solve it with dlnetwork & Deep Learning Custom Training Loops based approach.