Microsoft has unveiled ML.NET 3.0, the latest iteration of its open-source, cross-platform machine learning framework designed to seamlessly integrate machine learning models into .NET applications.
This release, announced on November 27, introduces novel deep learning capabilities, including advancements in object detection, named entity recognition, and question answering.

ML.NET 3.0 incorporates these deep learning features through integrations and interoperability with TorchSharp and ONNX models. Additionally, the update enhances integration with the LightGBM gradient boosting framework. Developers can access ML.NET 3.0 from dotnet.microsoft.com.

Key Features of ML.NET 3.0:

1. Object Detection: Microsoft introduced Object Detection in ML.NET Model Builder in May, based on TorchSharp-powered Object Detection APIs in ML.NET 3.0. The Object Detection API leverages cutting-edge techniques from Microsoft Research, supported by a Transformer-based neural network architecture.
2. Natural Language Processing (NLP): ML.NET 3.0 expands its capabilities in NLP with features such as question answering and named entity recognition. These functionalities build upon the existing TorchSharp RoBERTa text classification features introduced in ML.NET 2.0.
3. Automated Machine Learning (AutoML): The release introduces new AutoML capabilities, with the AutoML Sweeper now supporting sentence similarity, question answering, and object detection.
4. DataFrame Enhancements: ML.NET 3.0 improves support for data processing scenarios, offering enhancements and bug fixes to DataFrame. Notable updates include expanded data loading from and exporting to SQL databases via ADO.NET, improved arithmetic performance in column cloning and binary comparison scenarios, and enhanced handling of null values during arithmetic operations.

As Microsoft looks ahead to .NET 9 and ML.NET 4.0, users can anticipate updates to Model Builder and the ML.NET CLI to accommodate the ML.NET 3.0 release. Future plans involve expanding deep learning scenarios, enhancing DataFrame functionalities, and integrating additional APIs into ML.NET, particularly those in System.Numerics.Tensors. Microsoft remains committed to advancing the capabilities of ML.NET to empower developers in the evolving landscape of machine learning and .NET applications.