liblinear-ruby

Liblinear-Ruby provides a Ruby interface to LIBLINEAR, a popular library for large-scale linear classification and regression. It enables Ruby developers to leverage efficient machine learning algorithms for tasks like support vector classification and logistic regression directly within Ruby applications.

Key Features

• LIBLINEAR Integration — Full Ruby bindings to LIBLINEAR 2.30 using SWIG for seamless interoperability.
• Multiple Solver Types — Supports various solvers including L2-regularized logistic regression, support vector classification (primal/dual), and support vector regression.
• Flexible Training Data — Accepts training data as arrays of arrays or arrays of hashes for feature representation.
• Model Persistence — Save trained models to files and load them later for prediction without retraining.
• Cross Validation — Built-in k-fold cross validation to evaluate model performance reliably.
• Feature Weights Access — Retrieve feature weights from trained models to analyze feature importance.
• Quiet/Verbose Modes — Control output verbosity during training for cleaner logs in production.

Philosophy

Liblinear-Ruby aims to make LIBLINEAR's powerful linear classification and regression algorithms accessible to Ruby developers with a clean, idiomatic Ruby API while maintaining full compatibility with the underlying C++ library.