using predictAndUpdateState with an LSTM with featureInputLayer

Hello,

I'd like to ask if it's possible to use "predictAndUpdateState" with an LSTM that has an featureInputLayer?

If I train my LSTM with a sequenceInputLayer everything works fine, but when I use a featureInputLayer the training works, prediction using the regular predict() works as well, but for "predictAndUpdateState" I get the error:

predictAndUpdateState and classifyAndUpdateState support recurrent neural networks only.

Is it possible to make predictAndUpdateState work using an LSTM that has been trained using a featureInputLayer?

Answer 1

[Zhiyu WANG]

LSTM requires somelike "timeseries" data while "featureInputLayer" is not along with that.Yo ucan refer to this answer:

https://ww2.mathworks.cn/matlabcentral/answers/1594784-when-use-sequence-and-when-feature-for-neural-networks-input-layer-in-regression-problem#answer_840364