how to evaluate LSTM network

hi, this is general question. how can i evaluate LSTM network in regression problem? i know from RMSE but after i train my net the accuracy on the training option becoma N/A? is there any function other than RMSE to evaluate my network especially in regression problem?

Answer 1

[Shashank Gupta]

Hi Safwana,

RMSE is generally the most typical loss function to use in regression problem and most of the time it gives optimal results. if you not happy with the loss function you can consider Mean Absolute error in place of MSE as well. But I am not sure whether it will improve your result or not. Also if you need help in evaluating your model performance after training, you can refer to this link.

From my first guess about RMSE loss showing N/A is probably because you are looking at validation or testing RMSE and you might not have provided data for validation or testing during the training of network. If the validation data is not provided the RMSE for validation will be shown as N/A. check out the data distribution properly. I am sure you will be able to figure it out.

Cheers.