Microsoft has released the first developer preview of ML.NET, their cross-platform open source machine framework for .NET developers. It was in development at Microsoft research and has been used in a variety of products including Windows, Azure, Bing, etc.
For developers without prior expertise in dealing with ML models, Microsoft says, the framework helps them in developing and adding custom ML to apps using their .NET and C# skills. Being cross-platform, it can run on Windows, Linux, and macOS devices.
The framework can be extended to add support for well known deep learning libraries like TensorFlow, Caffe2, and CNTK, and also for general machine learning libraries like Accord.NET.
ML.NET preview comes with core framework components including learning algorithms, transforms, and core ML data structures. It also brings the first draft of .NET APIs for training models and using models for predictions.
The preview enables ML tasks like classification (e.g., sentiment analysis and text categorization) and regression (e.g., forecasting and price prediction). Over time, it will enable other ML tasks like anomaly detection, recommendation system, and other approaches like deep learning using the benefits of added libraries.
In the coming future, Microsoft has plans to add support for the industry-standard open source deep learning model format ONNX. Further, there will be a better user interface to simplify machine learning tasks and the framework will be integrated with VS tools for AI.
You can find more information about ML.NET on its GitHub repo and on Microsoft’s website.
Check out more interesting stuff from Build 2018.