DISTANCE METRIC LEARNING BY QUADRATIC PROGRAMMING BASED ON EQUIVALENCE CONSTRAINTS

This is a toolbox that implements the method described in our paper "Distance Metric Learning by Quadratic Programming Based on Equivalence Constraints". All programs are written in Matlab, and you can also find the implementation of the method of Tsang and Kwok ("Distance Metric Learning with Kernels" in Int. Conf. on Artificial Neural Networks). We revised the codes such that the resulting distance matrix is always positive semi-definite (The authors claim that the resulting distance matrix is positive semi-definite in their paper, but this is incorrect).

Install:

• Download DMLQP.zip files and uncompress them in an appropriate directory.
• Add the toolbox directory in the Matlab path.
• Launch the demo programs in order to verify if the toolbox works properly, the program takes only a few minutes to finish.
• Please read the function descriptions and explanations in demo programs in order to learn their usage. Please cite the paper "Distance Metric Learning by Quadratic Programming Based on Equivalence Constraints" by Hakan Cevikalp if you used this toolbox.

If you find a bug, please send an email to hakan.cevikalp@gmail.com