Getting best fitting model when using trainlm

I am running a script which uses the 'trainlm' neural net training algorithm multiple times. It time that it runs it stops when it has not improved for 7 epochs. It then, it seems, provides the model for that last epoch (epoch 'n'), not the best fitting model from epoch n-7, which often is substantially better. Is this correct and is there any way to get the model from the best fitting epoch?

Answer 1

[Greg Heath]

Early stopping ONLY depends on the 15% validation subset performance. NOT on the 70% training or 15% testing performance.

Increasing val set error is proof that the net is not generalizing well to nontraining data.

Since the decreasing improvement neither refers to the 70% training or 15% testing subset, there is no proof that there is a need to back up.

Hope this helps.

Thank you for formally accepting my answer

Greg

[Walter Johnson]

Thanks Greg. That is clear and should have been obvious.