how to train LSTM with single input and two outputs?

hello everyone,

I have question regarding the training of LSTM network. I want to train my network with 1 input and 2 outputs.

Network architecture is as:

layers = [ ...

sequenceInputLayer(numFeatures,'Normalization', 'zscore')

lstmLayer(numHiddenUnits,'OutputMode','sequence')

lstmLayer(numHiddenUnits,'OutputMode','sequence')

lstmLayer(numHiddenUnits2,'OutputMode','sequence')

lstmLayer(numHiddenUnits2,'OutputMode','sequence')

fullyConnectedLayer(numResponses)

regressionLayer];

with numFeatures=1 and numResponses=2.

Do i have to make custom regression layer for 2 output as i read that for multiple input and single output, custom regression layer is needed to train the network but there is no information for multiple out.

anybody can help me in this regard.

Thanks.

Answer 1

[Prateek Rai]

To my understanding, you want to train LSTM with two outputs.

You can refer to MATLAB Answer on LSTM Example for Multi input and Multi outputs to get more idea.