Is a LSTM parameter to sequence regression possible?

Hello,

What happens if I have for example 30 different input parameters in a dataset and a corresponding signal as output and I want to predict this signal?

E.g. features are [X1, X2, X3, .... X30] and the label is a time dependent signal of length n [X31(t_1) X(31(t_2) X(31(t_3) .... X31(t_n)]

layers = [ ...

fullyConnectedLayer(30)

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

fullyConnectedLayer(n)

regressionLayer];

This did not work for me so far as I think there is a problem with the input layer?

Can someone help?

Answer 1

[Divya Gaddipati]

For a sequence input, you can use sequenceInputLayer.

sequenceInputLayer(featureDimension)

For more informatiom on sequenceInputLayer, refer to the following link:

https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.sequenceinputlayer.html

Here's an example on Sequence-to-Sequence regression:

https://www.mathworks.com/help/deeplearning/ug/sequence-to-sequence-regression-using-deep-learning.html