wolffd@0: function [e, edata, eprior] = errbayes(net, edata)
wolffd@0: %ERRBAYES Evaluate Bayesian error function for network.
wolffd@0: %
wolffd@0: %	Description
wolffd@0: %	E = ERRBAYES(NET, EDATA) takes a network data structure  NET together
wolffd@0: %	the data contribution to the error for a set of inputs and targets.
wolffd@0: %	It returns the regularised error using any zero mean Gaussian priors
wolffd@0: %	on the weights defined in NET.
wolffd@0: %
wolffd@0: %	[E, EDATA, EPRIOR] = ERRBAYES(NET, X, T) additionally returns the
wolffd@0: %	data and prior components of the error.
wolffd@0: %
wolffd@0: %	See also
wolffd@0: %	GLMERR, MLPERR, RBFERR
wolffd@0: %
wolffd@0: 
wolffd@0: %	Copyright (c) Ian T Nabney (1996-2001)
wolffd@0: 
wolffd@0: % Evaluate the data contribution to the error.
wolffd@0: if isfield(net, 'beta')
wolffd@0:   e1 = net.beta*edata;
wolffd@0: else
wolffd@0:   e1 = edata;
wolffd@0: end
wolffd@0: 
wolffd@0: % Evaluate the prior contribution to the error.
wolffd@0: if isfield(net, 'alpha')
wolffd@0:    w = netpak(net);
wolffd@0:    if size(net.alpha) == [1 1]
wolffd@0:       eprior = 0.5*(w*w');
wolffd@0:       e2 = eprior*net.alpha;
wolffd@0:    else
wolffd@0:       if (isfield(net, 'mask'))
wolffd@0:          nindx_cols = size(net.index, 2);
wolffd@0:          nmask_rows = size(find(net.mask), 1);
wolffd@0:          index = reshape(net.index(logical(repmat(net.mask, ...
wolffd@0:             1, nindx_cols))), nmask_rows, nindx_cols);
wolffd@0:       else
wolffd@0:          index = net.index;
wolffd@0:       end
wolffd@0:       eprior = 0.5*(w.^2)*index;
wolffd@0:       e2 = eprior*net.alpha;
wolffd@0:    end
wolffd@0: else
wolffd@0:   eprior = 0;
wolffd@0:   e2 = 0;
wolffd@0: end
wolffd@0: 
wolffd@0: e = e1 + e2;