Accessing the LSTM internal state

As already written above, I am interested in accessing the internal state of a built-in LSTM layer, which is just one layer of a bigger network I have to implement for time series classification. Unfortunately, it seems MATLAB provides initial conditions only for both cell and hidden states, but what about retrieving them at each running step? Is collecting the time varying states possible in this context?

Answer 1

[Giuseppe Dell'Aversana]

Hi! Have you found an answer?