LSTM - Way to set inputweight based on user defined variable

Dear everyone,

I am really interested in LSTM to use predicting the time series data

In the progress, I want to set inputweight based on user defined value

I thought that the size of inputweights should be 'InputSize', however, it should be '4*NumHiddenUnitsxInputSize' as mentioned in

https://kr.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.lstmlayer.html

I tried to find the reason that size of inputweights is '4*NumHiddenUnitsxInputSize', but I cannot find it

If there are someone who know the reason that the size of inputweights is '4*NumHiddenUnitsxInputSize', please explain it to me..

Thank you

Sincerly yours

Answer 1

[Srivardhan Gadila]

You can find the explanation in the same documentation page in More About section.

You can also refer to Long Short-Term Memory (LSTM) & LSTM Layer Architecture.