Machine Learning .NET 2.0 enhances text classification

Published on Nov 24, 2022

An upgrade to Microsoft’s machine learning framework for .NET improves text classification model building, introduces a similarity API, and extends AutoML’s capabilities.

The open source, cross-platform machine learning framework for .NET, ML.NET 2.0, has been released by Microsoft. Automated machine learning and text classification capabilities are included in the upgrade.

The ML.NET 2.0 was announced on November 10 along with an updated version of the ML.NET Model Builder, a visual development tool for building machine learning models for .NET applications. The Model Builder presents a text classification scenario powered by the ML.NET Text Classification API.

Developers can train custom models to classify raw text data using the Text Classification API, which was showcased in June. To fine-tune the model, the Text Classification API uses a pre-trained TorchSharp NAS-BERT model from Microsoft Research. Model Builder supports local training on either CPUs or GPUs that are compatible with CUDA.

In ML.NET 2.0, you will also find:

Preconfigured automated machine learning pipelines make it easier to start using machine learning for binary classification, multiclass classification, and regression models.

The AutoML Featurizer can be used to automate data preprocessing.

A developer has the option of selecting which trainers to use as part of a training process. Additionally, they can select the tuning algorithms that will be used to find the optimal hyperparameters.

An advanced AutoML training option is introduced that allows the user to choose trainers and an evaluation metric to optimize.

With the same underlying TorchSharp NAS-BERT model, a sentence similarity API calculates a numerical value representing similarity between two phrases.

ML.NET’s future plans include expanding deep learning coverage and emphasizing the use of the LightBGM framework for classical machine learning tasks such as regression and classification. Additionally, the developers of ML.NET intend to enhance the AutoML API in order to facilitate new scenarios and customizations, as well as simplify the machine learning workflows.