Can we customize loss function for Neural State Space Model?

Hello,

I am trying to train my system with Neural State Space Model. Is there a way to customize the loss function instead of using MSE or MAE?

Thank you.

Answer 1

[NVSL]

Hi @Saskia Putri

To customize your training process by defining a custom loss function and updating the learnable parameters, you can use the "dlupdate" function.

For a clearer understanding on usage of dlupdate in training loop, you can refer to the example provided in the documentation:

https://www.mathworks.com/help/deeplearning/ref/dlupdate.html#mw_fe0c6eb9-f0e2-45f7-bd97-8327fd580d17

For more detailed information on modelling loss function for custom training loop, please refer to the following documentation:

https://www.mathworks.com/help/deeplearning/ug/define-model-gradients-function-for-custom-training-loop.html