At this year’s Build conference, Microsoft announced the debut release of ML.NET. ML.NET is an open source, cross-platform, machine learning framework for .NET. Microsoft’s Ankit Asthana has now announced that the project has made its second release. This edition adds several new features including a new machine learning task called clustering, cross-validation and train-test, and a new repo on GitHub demonstrating how to use both the new and existing features.
One of the changes made in ML.NET 0.2 is its ability to support loading data sets from a collection of objects, previously these could only be loaded from a delimited text file. Another addition is cross-validation, which is a method of validating the performance of a machine learning model. A useful aspect of the cross-validation approach is that it does not require a dataset separate from what was used to create the model. Instead, it partitions the provided data multiple times into different groups of training and testing data.
Joining the release of ML.NET 0.2 is a growing repository of sample code that demonstrates how to use this new framework. Categories range from basic examples to learn about the new concepts, to larger applications that are full feature demonstrations of applying the technology.
ML.NET supports projects being developed on Linux, MacOS, and Windows. Those looking to get started with ML.NET can easily do so by adding the ML.NET NuGet package to their project in either Visual Studio or from the command line. Microsoft has provided full release notes, and a comprehensive list of the 36 items fixed in 0.2.