MATLAB: Combining two neural networks (net1 is trained & net2 is untrained) in one bigger network

Deep Learning ToolboxMATLABneural networks

I have two shallow neural network models net1, net2.

I have a trained net1. Then I need to connect a net2 to a net1. Hence I treat the combined (net2+net1) as one large net and then train the big net . However, during the second training phase, we only need to update the weights corresponding to net2 (and keep net1 weights fixed). Is it possible in MATLAB? any suggested links or articles to perform this part.

Best Answer

How about this?:

https://www.mathworks.com/matlabcentral/answers/497749-avoid-training-certain-neurons#answer_407422?s_tid=prof_contriblnk

Related Solutions

MATLAB: Net reusing

Inside your loop, the input to train() is definitely the value of "net" that just calculated in the previous. I am guessing you want to do something like this instead:

net0 = patternnet(10);
% net0 initialization
% ...
for ii=1:10
   net = train(net0,in,tar);
   netSet{ii} = net;
end

This way, you are starting from the initialized value "net0" each time. I don't know enough about neural nets to know if train() has some randomness in it that will make each value of netSet(ii) different.

Also, note that you do not really need the intermediate value "net" inside your loop. You could just do

for ii=1:10
   netSet{ii} = train(net0,in,tar);
end

MATLAB: Neuronal Net training

Which network?

Which outputs?

help train

doc train

Look at the explanation of what outputs are available at the END of training; e.g.,

[net tr Y E Pf Af ] = train(net,p,t,Pi,Ai );

The only information on what happened DURING training is in the structure tr and the command line output. Try

net.trainParam.show= 1;

If you want to obtain a complete history of [net ... Ai], try removing the semicolon.

WARNING: If you try to obtain a history by looping over a succession of 1 epoch training calls, you can get a valid result. HOWEVER, you will not get the same result as that from a single multi-epoch call.

Hope this helps.

Greg

Best Answer

Related Solutions

MATLAB: Net reusing

MATLAB: Neuronal Net training

Related Question