Microsoft Corporation at its three day build conference held in Seattle, Washington announced the preview release of a machine learning framework called ML.NET. Developed by the research subsidiary, Microsoft Research, the framework will assist .NET developers in developing their own models for their web apps across Windows, Linux and macOS platform. Developers can infuse the custom machine learning models into applications without much prior experience in building machine learning models.
The current release 0.1 is the debut preview compatible with any of the platforms that support .NET Core 2.0 or .NET Framework.
Developers can access the framework directly from Github.
Apart from the machine learning capabilities, this debut preview of ML.NET also uncovers draft of .NET APIs schemed for developing models for prediction, and training of machine learning models, different machine learning algorithms and core ML data structures.
Although it is the first release, Microsoft and its team have been using this framework in their various product groups like Azure, Bing and Windows.
Microsoft has also mentioned clearly that soon, ML.NET will include more advanced machine learning scenarios such as recommendation systems and anomaly detection. Popular concepts like deep learning, and support for libraries like TensorFlow, CNTK, and Caffe2 would be added. Support for general machine learning libraries like Accord.NET framework would also be included in the near soon release. The framework would also add miscellaneous support to ONNX, scaling out on Azure, Better GUI for ML tasks simplification and integration support with VS Tools.
To follow the progress on this framework, visit .NET Blog on Microsoft’s official site.