Infer.NET

Infer.NET, originally developed by Microsoft Research's Machine Intelligence and Perception group and now published as open source under the .NET Foundation (MIT license), is a framework for Bayesian inference in graphical models — supporting probabilistic programming for classification, recommendation, clustering, and domain-specific applications in fields like bioinformatics and computer vision. It ships built-in learners including Bayes Point Machine classifiers and a Matchbox recommender, giving developers pre-built solutions for common probabilistic ML problems rather than requiring every model to be built from first principles. As a specialized Bayesian-inference framework, it's aimed at developers and researchers already familiar with probabilistic modeling, not a general-purpose ML tool. It's free, available on GitHub and as NuGet packages, with an archived Q&A community and active GitHub issue tracking for support. Good fit for a .NET developer or researcher who needs Bayesian inference and probabilistic programming specifically, rather than standard supervised/unsupervised ML.