Wolfram Language

Explore Training Results

This example demonstrates how to use a NetTrainResultsObject to programmatically explore performance measurement histories, timing and efficiency metrics and other results collected while training a net, as well as how to use NetTrain to record custom training measures.

Train LeNet on FashionMNIST.

Examine the final plots.

Compare the final and best validation measurements.

Compare the final loss on the training and validation sets.

Query various training efficiency metrics.

Get a list of all of the available properties.

If the preceding properties are not enough, it is possible to collect custom measures such as the evolution of the per-layer gradient magnitudes, the per-example losses or the learning rate during the training of a net.

Obtain a plot of the evolution of the per-array gradient magnitudes during the training of a LeNet on FashionMNIST.

Record the losses of individual examples over time.

Plot the loss associated with a single example over time.

Find the most difficult examples by calculating the mean loss for each example and taking the indices of the 20 largest such mean losses.

Show the average evolution of losses and error rates for bags, coats, dresses and shirts.

Related Examples