dotnet/machinelearning
Machinelearning
This is a cross-platform framework for building, training, and deploying custom machine learning models within the .NET ecosystem. It provides a predictive modeling engine for classification, regression, and forecasting tasks, alongside an inference runtime to generate predictions across different hardware architectures.
The framework includes a gradient boosting library and supports interoperability with external models via a standardized open format. It features tools for prediction explainability, allowing the analysis of feature importance to debug model behavior and identify bias.
The project covers the full machine learning lifecycle, including data transformation pipelines for preprocessing tabular data, deep learning model execution for entity recognition and object detection, and native analytics kernels to accelerate training and inference.
The repository includes command-line scripts for dependency restoration and batch compilation of its source code.
Features
- .NET Machine Learning Integrations - Integrates predictive models and machine learning capabilities directly into applications built with the .NET framework.
- Cross-Platform Inference Frameworks - Provides a runtime engine that executes trained models across diverse architectures, including ARM64 and WebAssembly.
- Custom Predictive Model Development - Enables the development of custom classification and forecasting models using built-in algorithms and transformation pipelines.
- Training - Implements classification and regression tasks using an optimized gradient boosting framework.
- Machine Learning Frameworks - Provides a comprehensive cross-platform framework for building, training, and deploying custom machine learning models in .NET.
- Model Deployment - Supports loading and executing pre-trained machine learning models from external formats within software applications.
- Model Inference Runtimes - Provides an execution engine for running trained models and external pre-trained formats across different architectures.
- Model Predictions - Generates predictions and inferences from trained models when applied to new input data.
- Prediction Engines - Ships a toolset for training classification, regression, and forecasting models using tabular data and custom algorithms.
- Model Training - Enables developing and training classification or regression models using local datasets and built-in algorithms.
- Data Preprocessing Pipelines - Provides tools for cleaning and transforming raw datasets from files or databases to prepare them for ML pipelines.
- Data Transformation - Supports importing datasets from files and databases and applying transformations to prepare raw information for training.
- Data Transformation Pipelines - Provides sequential data transformation pipelines to prepare raw datasets for model training and inference.
- Tabular Data Processors - Provides a tabular interface to filter, merge, and transform datasets for machine learning pipelines.
- Deep Learning Inference Engines - Includes runtime engines for executing deep learning models such as transformers for object detection and entity recognition.
- Gradient Boosting Libraries - Includes a library for implementing classical machine learning tasks using optimized gradient boosting algorithms.
- Feature Importance Attribution - Includes tools for analyzing the contribution of individual input variables to model predictions for debugging and transparency.
- External Model Loading - Supports importing and executing pre-trained models from standardized open formats to extend algorithmic capabilities.
- Model Interoperability Formats - Supports loading and executing machine learning models from the standardized ONNX open format for cross-framework compatibility.
- Model Explainability - Analyzes feature importance and contributions to interpret model decisions and identify potential bias.
- Dynamic Schema Inference - Automates preprocessing and training by dynamically inferring input and output shapes from the provided data.
- Native Analytics Kernels - Uses optimized low-level mathematical kernels to accelerate the execution of training algorithms and inference.
- Schema Inference - Automatically determines input and output data shapes at runtime to support datasets without predefined structures.
- Cross-Platform Runtimes - Provides a unified execution layer to run trained models across diverse hardware architectures including ARM64 and WebAssembly.
- Custom Kernel Accelerators - Implements optimized low-level mathematical kernels to accelerate the execution of training algorithms and data processing.
- General Machine Learning - Cross-platform machine learning framework for .NET developers.
- Machine Learning and Data Science - Cross-platform framework for building machine learning models.
- Model Conversion Tools - Framework for integrating machine learning into .NET applications.
- Developer Utilities - Library for integrating predictive models into applications.
