LIBSVM
| Company type | Open-source software |
|---|---|
| Industry | Machine learning software |
| Founder | National Taiwan University |
| Products | Machine learning toolkits (support vector machines, logistic regression) |
LIBSVM and LIBLINEAR are well-known open-source machine learning toolkits created at National Taiwan University. Both are implemented in C++ and provide a C interface. LIBSVM employs the sequential minimal optimization (SMO) technique to train kernel-based support vector machines (SVMs), enabling both classification and regression tasks. In contrast, LIBLINEAR specializes in linear SVMs and logistic regression models, utilizing a coordinate descent optimization method for training.
The SVM algorithms from these libraries have been incorporated into multiple open-source machine learning frameworks, including GATE, KNIME, Orange, and scikit-learn. They support various programming languages through dedicated bindings such as Java, MATLAB, R, Julia, and Python. Notably, LIBSVM is accessible via R's e1071 package and integrated within Python's scikit-learn library.
Both LIBSVM and LIBLINEAR are distributed as free software under the terms of the 3-clause BSD license.