ML.NET 2.0 improves text classification

Microsoft has released ML.NET 2.0, a new version of its open-source cross-platform machine learning framework for .NET. The upgrade includes text classification and automatic machine learning features.

Unveiled November 10, ML.NET 2.0 arrived in tandem with a new version of the ML.NET Model Builder, a visual development tool for creating machine learning models for .NET applications. The model builder introduces a text classification scenario powered by the ML.NET Text Classification API.

Preview in June, the Text Classification API allows developers to train custom models to classify raw text data. The Text Classification API uses a pre-trained TorchSharp NAS-BERT model from Microsoft Research and the developer’s own data to refine the model. The Model Builder scenario supports local training on CUDA-enabled CPUs or GPUs.

Also in ML.NET 2.0:

• Binary classification, multiclass classification, and regression models using preconfigured machine learning pipelines make it easy to use machine learning.
• Data pre-processing can be automated using the AutoML Featureizer.
• Developers can choose which trainers to use as part of a training process. They can also choose tuning algorithms used to find optimal hyperparameters.
• Advanced training options on AutoML are introduced to choose trainers and choose an evaluation metric to optimize.
• A sentence similarity API, using the same underlying TorchSharp NAS-BERT model, calculates a numerical value representing the similarity of two sentences.

Future plans for ML.NET include expanding deep learning coverage and emphasizing the use of the LightBGM framework for classic machine learning tasks such as regression and classification. The developers behind ML.NET also intend to improve the AutoML API to enable new scenarios and customizations and simplify machine learning workflows.