Microsoft has open sourced the Infer.NET, a machine learning engine that the company uses to power its own platforms including Office, Azure, and Xbox.
Infer.NET is now available on GitHub under the permissive MIT license. Enterprises can use it for free for their commercial applications.
Launched in 2004 as a research tool, the Infer.NET has evolved over the years to become a successful machine learning framework.
Microsoft has built the tool to enable a model-based approach, allowing users to integrate domain-based knowledge into the model. The company explained that Infer.NET can then build a bespoke machine learning algorithm directly from the model.
Typically, the conventional machine learning tools provide existing AI algorithms, so that users can push them into the project according to the needs. But, Infer.NET is different. It constructs the machine learning algorithm on the basis of model provided by the users.
“An added advantage of model-based machine learning is interpretability. If you have designed the model yourself and the learning algorithm follows that model, then you can understand why the system behaves in a particular way or makes certain predictions,” wrote Yordan Zaykov, Principal Research Software Engineering Lead, in a blog post.
“As machine learning applications gradually enter our lives, understanding and explaining their behavior becomes increasingly more important.”
The models built using Infer.NET are capable of handling a broad range of data traits, like real-time data, heterogenous data, unlabeled data, insufficient data, etc.
Microsoft will make Infer.NET a part of ML.NET, the machine learning framework for .NET developers. The tech giant has already taken a number of steps to integrate it with ML.NET, like setting up the repository under .NET Foundation and moving the packages and namespaces to Microsoft.ML.Probabilistic.