Extended Fisher Discriminant Analysis and Point-Ellipsoid Distance

Extended FDA for binary classification and Point-Ellipsoid Distance

This submission contains functions for running the extended Fisher Discriminant Analysis (eFDA.m) for binary classification given a training (and optionally test) dataset. Under the hood the method requires the calculation of the distance between a point and an ellipsoid (in arbitrary dimention), and this is provided as a separate code pointellipsoiddistance.m to allow its independent use.

Details on the underlying algorithms can be found in
'Global Optimization methods for extended Fisher Discriminant Analysis',
by S. Iwata, Y. Nakatsukasa, and A. Takeda, accepted to AISTAT 2014.

For details of eFDA, see
"A unified classification model based on robust optimization."
by A. Takeda, H. Mitsugi, and T. Kanamori, Neural Computation, 25(3):759-804,2013.