Measure the Quality of a Net

This example demonstrates how to measure the performance of a net both during training (using TrainingProgressMeasurements) and after training (using NetMeasurements).

Train "LeNet" on a sample of "FashionMNIST" and examine the F1 score, accuracy and confusion matrix during training (click the following image to see an animation).

Measure the accuracy of the trained net.

Measure the F1 score for each class.

Measure the macro-averaged precision and recall.

Examine the confusion matrix.

Examine the mean activations for the filters in the first convolution layer.