News

Microsoft Open Sourcing Infer.NET Framework, Will Integrate with ML.NET

• By Becky Nagel
• 10/08/2018

On Friday, Microsoft announced that Infer.NET, a cross-platform framework for model-based machine learning, is now open source and available on GitHub.

With model-based machine learning, instead of providing a number of models that the data may (or may not) fit into, the goal is instead to provide a framework that can support a wide variety of data models that can be tailored for one particular project. (Microsoft offers a free book explaining model-based machine learning here.)

And that's how Infer.NET works, according to Microsoft Research, which started the project back in 2004. The framework builds a custom machine learning algorithm directly from that model you provide: "This means that instead of having to map your problem onto a pre-existing learning algorithm that you've been given, Infer.NET actually constructs a learning algorithm for you," the company explained.

Several aspects of Infer.NET set it apart, Microsoft Research states, including its use of probabilistic programming, deterministic inference algorithms and the way it supports online Bayesian inference.

Like machine learning itself, the platform has transformed from primarily research-based uses cases to more and more commercial applications. Even so, it's still regularly used in academic projects (a list of papers can be found here).

According to Microsoft, along with becoming open source, Infer.NET will no longer stand alone: "Infer.NET will become a part of ML.NET -- the machine learning framework for .NET developers. We have already taken several steps towards integration with ML.NET, like setting up the repository under the .NET Foundation and moving the package and namespaces to Microsoft.ML.Probabilistic. Infer.NET will extend ML.NET for statistical modeling and online learning."

"The Infer.NET team is looking forward to engaging with the open-source community in developing and growing the framework further," Microsoft Research continued.