/CompositeQuantization

Primary LanguageC++GNU General Public License v2.0GPL-2.0

Composite Quantization

What is it?

This software library implements the composite quantization algorithm described in

Composite Quantization for approximate nearest neighbor search. Ting Zhang, Chao Du and Jingdong Wang. In International Conference on Machine Learning (ICML), 2014.

If you use this software for research purposes, please cite the aforementioned paper in any resulting publication.

The latest version

Version 1.0 (2015-01-28): Initial release.

Installation

See HOW_TO_INSTALL.txt.

Documentation

See HOW_TO_TRAINING.txt and HOW_TO_SEARCH.txt.

Example usage

See demo.cpp and config.txt in the source code.

Reference

libLBFGS: http://www.chokkan.org/software/liblbfgs/index.html

Fast k-means: http://research.microsoft.com/en-us/um/people/jingdw/LargeScaleClustering/index.html

Product quantization: http://people.rennes.inria.fr/Herve.Jegou/projects/ann.html