hybrid_lda package
Author: Brian D. Bue (bbue@rice.edu)

----- Description -----

Implementation of hybrid metric learning algorithm described in:

B. Bue and E. Merényi, "An Adaptive Similarity Measure for
Classification of Hyperspectral Signatures," IEEE Geoscience and
Remote Sensing Letters, 2012.

Please cite the reference above if you use this code in a publication.

----- Example usage -----

To run the Python demo for several different lambda regularization
values:

   lambdas = [1e-5,1e-3,0,0.1,0.5,0.9]
   minDistHybrid([dat_ci,dat_cr],labels,lambdas,Dlab=["CI","CR"])

where: 
- dat_ci and dat_cr: [N x d] arrays containing a set of N
  continuum-intact spectral signatures and their corresponding
  continuum-removed representations, respectively; 
- labels: N-dimensional vector of class labels for each signature; 
- lambdas: regularization values, each in the [0,1] range.

Functions to calculate the continuum-removed representation of a
spectral signature are provided in the LINCR library available at:
http://www.ece.rice.edu/~bdb1/#code.

Tested on: 
- OSX 10.6 
- Python 2.6, numpy 1.6, scipy 0.9
- Matlab R2011a

----- Copyright -----

Copyright 2010-2012 Rice University