get validation loss and metric for the best network

How to get the validation loss and metric for the best network.

Example code:

[net_dnn, net_info] = trainnet(inputs,targets,layers,'mse',options);

In the ValidationHistory, I do have the loss and the performance metric for every 50 iterations (ValidationFrequency was 50). As for MSE, I guess I could get the loss by finding the minimum value but is there any other way to get that?

Answer 1

[Matt J]

You would have to interpolate the validation history data at the best MSE, probably.

But keep in mind, the best MSE does not really correspond to the "best network". It only corresponds to the best minibatch.

[Md. Al-Imran Abir]

Thank you for the minibatch info. I guess I will need to predict on the validation data and then calculate the MSE.