The Schwartz Report

Blog archive

Microsoft Releases PowerShell Module for Azure Machine Learning

Microsoft last week released the preview of the PowerShell Module for its Azure Machine Learning (ML) service. The Azure ML PowerShell cmdlets library, available on GitHub, lets users interact with Azure ML workspaces, experiments, Web services and endpoints.

The .NET-based PowerShell DLL module will let users fully manage their Azure ML workspaces, according to a blog post by Microsoft Principal Program Manager Hai Ning. The module includes the entire source code, which Ning said has a cleanly separated C# API layer.

"This means you can also reference this DLL from your own .NET project and operate Azure ML through .NET code," Ning stated. "In addition, you can also find the underlying REST APIs the DLL depends on and leverage the REST APIs directly from your favorite client as well."

The module will let IT pros and developers provision new ML workspaces through a management certificate, export and import JSON files that represent an experiment graph and to recreate new endpoints from published Web services, Ning added.

In a separate GitHub post, Ning described how to build a generic machine learning workflow using a global bicycle rental franchise looking to build a regression model to predict rental demand based on historical data. The post documents how to train the algorithm on multiple datasets with the same feature sets but using different feature values to produce models specific to each dataset using the Azure ML retraining API and Azure ML PowerShell automation.

Microsoft intends to extend the Azure ML PowerShell library, according to Ning. The Azure ML workspace and Web service are set to become managed resources with cmdlets pegged to support new ARM-based resources over time.

Posted by Jeffrey Schwartz on 05/09/2016 at 12:51 PM


comments powered by Disqus

Subscribe on YouTube