c@256: /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
c@256: 
c@256: /*
c@256:     QM DSP Library
c@256: 
c@256:     Centre for Digital Music, Queen Mary, University of London.
c@258:     This file copyright 2008 QMUL.
c@256:     All rights reserved.
c@256: */
c@256: 
c@256: #ifndef KLDIVERGENCE_H
c@256: #define KLDIVERGENCE_H
c@256: 
c@256: #include <vector>
c@256: 
c@256: using std::vector;
c@256: 
c@256: /**
c@258:  * Helper methods for calculating Kullback-Leibler divergences.
c@256:  */
c@256: class KLDivergence
c@256: {
c@256: public:
c@256:     KLDivergence() { }
c@256:     ~KLDivergence() { }
c@256: 
c@258:     /**
c@258:      * Calculate a symmetrised Kullback-Leibler divergence of Gaussian
c@258:      * models based on mean and variance vectors.  All input vectors
c@258:      * must be of equal size.
c@258:      */
c@258:     double distanceGaussian(const vector<double> &means1,
c@258:                             const vector<double> &variances1,
c@258:                             const vector<double> &means2,
c@258:                             const vector<double> &variances2);
c@258: 
c@258:     /**
c@258:      * Calculate a Kullback-Leibler divergence of two probability
c@258:      * distributions.  Input vectors must be of equal size.  If
c@258:      * symmetrised is true, the result will be the symmetrised
c@258:      * distance (equal to KL(d1, d2) + KL(d2, d1)).
c@258:      */
c@258:     double distanceDistribution(const vector<double> &d1,
c@258:                                 const vector<double> &d2,
c@258:                                 bool symmetrised);
c@256: };
c@256: 
c@256: #endif
c@256: