LIBSOL: Library for Scalable Online Learning

LIBSOL is an open-source library for scalable online learning algorithms, especially for high-dimensional data. It provides a family of regular and sparse online learning algorithms for binary and multi-class classification tasks, and is designed with high efficiency, scalability, portability and extensibility. LIBSOL provides easy-to-use command-line tools, python wrappers and library calls for users and developers, and comprehensive documents. LIBSOL is not only a machine learning toolbox, but also a comprehensive research platform for online learning . LIBSOL is released under the Apache 2.0 open source license.

Why LIBSOL?

Speed. Implemented in C++ and carefully optimized to reduce time and memory cost. We believe LIBSOL is one of the fastest implementations for large-scale online learning available (faster than VW and Liblinear).
Scalability. Data are stored in a sparse structure. All operations are optimized with sparse data structure.
Portability. All codes follow C++11 standard. No dependency on external libraries. Organized by CMake so that users on different platforms can build the library easily. LIBSOL can run on almost every platform.
Extensibility. Extensible code fosters active development. Easy to add new algorithms and new solutions.
Community: Join our LIBOL community on Github for more about Online Learning research.

Documentation

• Short Introduction
  A short paper for brief introduction of LIBSOL.
• Manual for Users & Developers
  Detailed manual and design documentation.
• WIKI Documentation
  Practical guide and framework reference on WIKI.
• Installation instructions
  Tested on Ubuntu, Red Hat, OS X.
• Tutorial Examples
  Run on data sets: a1a, mnist, rcv1, url, webspam, including comparisons to VW and LIBLINEAR.

Citing LIBSOL

Please cite the LIBSOL technical report in your publications if it helps your research:

@article{libsol2016,
  Author = {Wu, Yue and Hoi, Steven C.H. and Yu, Nenghai},
  Journal = {SMU Technical Report (SMU-TR-2016-07-25)},
  Title = {LIBSOL: A Library for Scalable Online Learning Algorithms},
  Year = {2016}
}