This software is made available under the Creative Commons 
Attribution-Noncommercial License.  You are free to use, copy, modify, and 
re-distribute the work.  However, you must attribute any re-distribution or 
adaptation in the manner specified below, and you may not use this work for 
commercial purposes without the permission of the author.

Any re-distribution or adaptation of this work must contain the author's name 
and a link to the software's original webpage.  This software comes with no 
guarantees, and all use of these codes is entirely at the user's own risk.

--------------------------------------------------------------------------------
Disentangling Factors of Variation via Manifold Interaction
--------------------------------------------------------------------------------

To run the code on Toronto Faces Database, first edit the file startup.m. If 
you are using jacket, specify JACKET_PATH. Otherwise, comment out that line and
the corresponding addpath(JACKET_PATH). Also specify LIBLINEAR_PATH and
TFD_PATH which contains TFD_48x48.mat. We have tested against liblinear-1.94.
 
Train using manifold objective (applied to emotion + ID):

>> script_tfd_man_man

Train using emotion labels and clamping ID units:

>> script_tfd_label_clamp

The classification and verification results will be saved in the log/
directory. The model files will be saved in the results/ directory.

--------------------------------------------------------------------------------

If you use this code for your research please cite:

Scott Reed, Kihyuk Sohn, Yuting Zhang, and Honglak Lee. "Disentangling
Factors of Variation via Manifold Interaction." Proceedings of The 31st
International Conference on Machine Learning. 2014.

