SomeB1oody/RustyML

What it solves

RustyML provides a comprehensive toolkit for machine learning and deep learning tasks entirely within the Rust ecosystem. It addresses the need for a native Rust library that handles data preprocessing, model training, and performance evaluation without requiring external language bindings.

How it works

The library is built on top of ndarray and utilizes Rust's native numeric kernels for computation. It offers a modular architecture where you can choose specific components—such as classical machine learning, neural networks, or evaluation metrics—via feature flags. Models can be saved and loaded using Serde and postcard for easy persistence, and it leverages Rayon to automatically parallelize heavy computations based on configurable work thresholds.

Who it’s for

This library is designed for Rust developers who want to implement machine learning models, neural networks, or graph-based architectures directly in their projects. It is suitable for those who prefer a pure Rust implementation and need fine-grained control over their dependencies and compilation footprint.

Highlights

• Supports classical machine learning, neural networks, and graph-based models.
• Includes built-in evaluation metrics for classification, regression, and clustering.
• Features modular compilation via feature flags to keep binary sizes small.
• Provides parallel processing for performance-critical operations.
• Enables model persistence through binary file saving and loading.
• Offers flexible graph construction for complex architectures like residual networks.