How can one integrate LSTM into the Neural State Space Deep Learning network. Also what is the exact difference between NSS vrs NeuralODE

Hello,

I have looked at the examples on using the neural state space and associated neural network used:MLPs.

I am interested in verify how an LSTM would perform differently for neural state space for system identification for prediction.

Could you help with how an LSTM can be integrated into the predefined Neural States Space framework. Since LSTM requires sequence input as opposed to feature input, an example on this done or custom version showing LSTM would be helpful.

Secondly, I want to understand the explicit difference between NNS and NeuralODE. I see both employ some ODE45 solver. I have read but not so clear on what is the difference. Lastly, how can I employ NeuralODE in Predictive control (MPC).

Thank you for your time