Validation and out-of-sample testing

Recovering losses and optimal parameters

To view the train and test loss on a particular model M, you simply invoke trainloss and testloss on the model.

To get the optimal theta and lambda recovered from training, you use thetaopt(M) and lambdaopt(M).

To recover the list of lambda values used for a regularization path, you simply use lambdapath(M). Similarly, the optimal thetas are found via thetapath(M), and you can find the corresponding training and test losses through trainloss(M) and testloss(M), respectively.