Summary

In this chapter, we set up our Azure Machine Learning workspace, created our first Azure Machine Learning compute cluster (Amlcompute) and ran an initial experiment on the cluster. Everything is automated, from the setup of the workspace, to cluster creation and the submission of the training script.

Azure Machine Learning helps you keep track of experiments, metrics, training scripts, logs, trained models, artifacts, metrics, images, and much more. In this chapter, we started from a small ML script and, step by step, added additional functionality to take advantage of modern ML infrastructure and management techniques. We registered experiments, executed runs locally and on the compute cluster, stored a snapshot of the training folder with each run, collected training scores per epoch and a test score per run, and streamed the output directly back to the notebook environment. With a few lines of code, you can compare the trained model with all previously registered models...