Demystifying the different Azure services for ML

Azure offers many services that can be used to perform ML – you can use a simple Virtual Machine (VM), a pre-configured VM for ML (also called Data Science Virtual Machine (DSVM)), Azure Notebooks using a shared free kernel, or any other service that gives you compute resources and data storage. Due to this flexibility, it is often very difficult to navigate through these services and pick the correct service for implementing an ML pipeline. In this section, we will provide clear guidance about how to choose the optimal ML and compute services in Azure.

First, it is important to discuss the difference between a simple compute resource, an ML infrastructure service, and an ML modeling service. This distinction will help you to better understand the following sections about how to choose these services for a specific use case:

• A compute resource can be any service in Azure that provides you with computing power, such...