Daniel@0: function g = demgpot(x, mix)
Daniel@0: %DEMGPOT Computes the gradient of the negative log likelihood for a mixture model.
Daniel@0: %
Daniel@0: %	Description
Daniel@0: %	This function computes the gradient of the negative log of the
Daniel@0: %	unconditional data density P(X) with respect to the coefficients of
Daniel@0: %	the data vector X for a Gaussian mixture model.  The data structure
Daniel@0: %	MIX defines the mixture model, while the matrix X contains the data
Daniel@0: %	vector as a row vector. Note the unusual order of the arguments: this
Daniel@0: %	is so that the function can be used in DEMHMC1 directly for sampling
Daniel@0: %	from the distribution P(X).
Daniel@0: %
Daniel@0: %	See also
Daniel@0: %	DEMHMC1, DEMMET1, DEMPOT
Daniel@0: %
Daniel@0: 
Daniel@0: %	Copyright (c) Ian T Nabney (1996-2001)
Daniel@0: 
Daniel@0: % Computes the potential gradient
Daniel@0: 
Daniel@0: temp = (ones(mix.ncentres,1)*x)-mix.centres;
Daniel@0: temp = temp.*(gmmactiv(mix,x)'*ones(1, mix.nin));
Daniel@0: % Assume spherical covariance structure
Daniel@0: if ~strcmp(mix.covar_type, 'spherical')
Daniel@0:   error('Spherical covariance only.')
Daniel@0: end
Daniel@0: temp = temp./(mix.covars'*ones(1, mix.nin));
Daniel@0: temp = temp.*(mix.priors'*ones(1, mix.nin));
Daniel@0: g = sum(temp, 1)/gmmprob(mix, x);