svm

Torch-SVM is a machine learning library for Torch7 that implements Support Vector Machine (SVM) algorithms. It provides efficient implementations of popular SVM methods, making it a valuable tool for researchers and developers working with classification tasks in the Torch ecosystem.

Key Features

• SGD-based SVM — Reimplementation of Leon Bottou's svmsgd and svmasgd algorithms for stochastic gradient descent training.
• LIBLINEAR wrapper — Interface to the well-known LIBLINEAR library for large-scale linear classification.
• LIBSVM interface — Mostly complete interface to LIBSVM for various SVM formulations (excluding precomputed kernel option).
• Torch7 integration — Seamlessly works with Torch7 tensors and data structures.

Philosophy

The project focuses on providing practical, production-ready SVM implementations that integrate smoothly with the Torch7 framework while maintaining compatibility with established SVM libraries.