Microsoft's ML.NET Machine Learning Framework Now a Release Candidate -- Visual Studio Magazine

Microsoft's ML.NET Machine Learning Framework Now a Release Candidate

By David Ramel
04/08/2019

Microsoft shipped its open source, cross-platform machine learning framework, ML.NET, as a Release Candidate, just one step away from general availability.

ML.NET helps .NET coders put their C# and F# skills to work in developing and integrating custom machine learning models into Web, mobile, desktop, gaming and Internet of Things (IoT) applications.

Just last month, Microsoft announced ML.NET 0.11, described as a stability release, with the promise of providing help to enterprises wishing to get it set up for production use.

That production-usage goal gets a step closer with the ML.NET 1.0 RC release, which finishes off changes to the main APIs. From here on, Microsoft said, the goal is to improve documentation and samples, and deal with any critical issues that arise -- all without introducing any breaking changes before general availability, expected soon.

"Soon we will be ending the first main milestone of a great journey in the open that started on May 2018 when releasing ML.NET 0.1 as open source," Microsoft said in an announcement last week. "Since then we’ve been releasing monthly, 12 preview releases so far."

Updates in the RC include:

Segregation of stable vs. preview version of ML.NET packages: Heading toward ML.NET 1.0, most of the functionality in ML.NET (around 95 percent) is going to be released as stable (version 1.0).
IDataView moved to Microsoft.ML namespace: Microsoft moved IDataView back into Microsoft.ML namespace based on developer feedback.
TensorFlow-support fixes: Some issues in v0.11 were addressed regarding support for TensorFlow, an open source machine learning framework used for deep learning scenarios (such as computer vision and natural language processing).

The full release notes detail all of the above and many more changes.

ML.NET is expected to hit general availability in the second quarter of the 2019 calendar year. As that includes the April-June timespan -- and releases have been cranked out in a regular monthly cadence going back a year -- developers should watch for the final product to debut next month.

And the company renewed its offer for help organizations take an ML.NET app into production, stating they can talk to an ML.NET engineer to:

Get help implementing ML.NET successfully in an application.
Provide feedback about ML.NET.
Demo an app and potentially have it featured on the ML.NET homepage, .NET Blog, or other Microsoft channel.

Interested organizations and developers can apply for that help here.

About the Author

David Ramel is an editor and writer at Converge 360.

Printable Format

comments powered by Disqus

Featured

Visual Studio Devs Demand Claude 3.5 Sonnet AI: 'Why Is VSC Always Preferred?'

The brand-new Visual Studio 2022 v17.12 lets devs specify the AI model they want to use with the baked-in GitHub Copilot, but some are clamoring for more options, such as the latest/greatest Claude 3.5 Sonnet model from Anthropic that is available in VS Code.
SQL Database in Microsoft Fabric in Preview for Faster AI Apps

Microsoft's early embrace of advanced AI in its dev tooling continues apace with the new public preview of SQL database in Microsoft Fabric, the company's analytics/data platform.
Microsoft Seeks to Sort & Simplify its Agentic AI Dev Story

The company will infuse multi-agent runtime tech developed in its AutoGen research project into its production-ready Semantic Kernel SDK, both open source.
Uno Platform Unveils 'Hot Design' Tool for Cross-Platform .NET Apps

Uno Platform provided a new take on visual designers for .NET app development, claiming a breakthrough that changes code at runtime rather than the design stage.
Simple k-NN Regression Using C#

Dr. James McCaffrey from Microsoft Research presents a complete end-to-end demonstration of k-nearest neighbors regression to predict a single numeric value. Compared to other machine learning regression techniques, k-NN regression is often slightly less accurate, but is very simple to implement and customize, and the results are highly interpretable.