wolffd@0: function [g, gdata, gprior] = glmgrad(net, x, t)
wolffd@0: %GLMGRAD Evaluate gradient of error function for generalized linear model.
wolffd@0: %
wolffd@0: %	Description
wolffd@0: %	G = GLMGRAD(NET, X, T) takes a generalized linear model data
wolffd@0: %	structure NET  together with a matrix X of input vectors and a matrix
wolffd@0: %	T of target vectors, and evaluates the gradient G of the error
wolffd@0: %	function with respect to the network weights. The error function
wolffd@0: %	corresponds to the choice of output unit activation function. Each
wolffd@0: %	row of X corresponds to one input vector and each row of T
wolffd@0: %	corresponds to one target vector.
wolffd@0: %
wolffd@0: %	[G, GDATA, GPRIOR] = GLMGRAD(NET, X, T) also returns separately  the
wolffd@0: %	data and prior contributions to the gradient.
wolffd@0: %
wolffd@0: %	See also
wolffd@0: %	GLM, GLMPAK, GLMUNPAK, GLMFWD, GLMERR, GLMTRAIN
wolffd@0: %
wolffd@0: 
wolffd@0: %	Copyright (c) Ian T Nabney (1996-2001)
wolffd@0: 
wolffd@0: % Check arguments for consistency
wolffd@0: errstring = consist(net, 'glm', x, t);
wolffd@0: if ~isempty(errstring);
wolffd@0:   error(errstring);
wolffd@0: end
wolffd@0: 
wolffd@0: y = glmfwd(net, x);
wolffd@0: delout = y - t;
wolffd@0: 
wolffd@0: gw1 = x'*delout;
wolffd@0: gb1 = sum(delout, 1);
wolffd@0: 
wolffd@0: gdata = [gw1(:)', gb1];
wolffd@0: 
wolffd@0: [g, gdata, gprior] = gbayes(net, gdata);