Deep Learning with Microsoft Cognitive Toolkit Quick Start Guide

Measuring performance when working with a manual minibatch loop

Using metrics when training with the regular APIs in CNTK is the easiest way to measure the performance of your model during and after training. Things will be more difficult when you work with a manual minibatch loop. This is the point where you get the most control though.

Let's first go back and review how to train a model using a manual minibatch loop. We're going to be working on the classification model we used in the section Validating performance of a classification model. You can find it in the Validating with a manual minibatch loop.ipynb file in the sample code for this chapter.

The loss for the model is defined as a combination of the cross-entropy loss function ...