This is the software package of our ICML paper:

@inproceedings{Bai2016Georeg,
 title={Differential Geometric Regularization for Supervised Learning of Classifiers},
 author={Qinxun Bai, Steven Rosenberg, Zheng Wu, and Stan Sclaroff},
 booktitle={International Conference on Machine Learning (ICML)},
 year={2016}
}

Run startup.m to add dependencies and compile mex files.

The main entry point is demo_RBF_CE.m, which is for cross-entrpy loss. For quadratic loss, just modify demo_RBF_CE.m by changing all the calls of "gradDist_RBF_CE" to "gradDist_RBF_Euclid".

To run the code, you also need to set up the directory of the data properly.
Please download the data from the following link
http://cs-people.bu.edu/qinxun/geo/geo.html
Suppose your root directory is called "root", then you should place the code and data like this:
root/code/
root/data/
then "cd root/code", and run startup.m
Finally, you can call, demo_RBF_CE('UCI','liver') to run tracker on dataset "liver" under the UCI repository. 

This code (excluding 3rd-party libraries) is released under the GPLv3 license for non-commercial use only. 
For other types of license please contact authors.
qinxun.bai@gmail.com