wolffd@0: function net = netunpak(net, w)
wolffd@0: %NETUNPAK Separates weights vector into weight and bias matrices. 
wolffd@0: %
wolffd@0: %	Description
wolffd@0: %	NET = NETUNPAK(NET, W) takes an net network data structure NET and  a
wolffd@0: %	weight vector W, and returns a network data structure identical to
wolffd@0: %	the input network, except that the componenet weight matrices have
wolffd@0: %	all been set to the corresponding elements of W.  If there is  a MASK
wolffd@0: %	field in the NET data structure, then the weights in W are placed in
wolffd@0: %	locations corresponding to non-zero entries in the mask (so W should
wolffd@0: %	have the same length as the number of non-zero entries in the MASK).
wolffd@0: %
wolffd@0: %	See also
wolffd@0: %	NETPAK, NETFWD, NETERR, NETGRAD
wolffd@0: %
wolffd@0: 
wolffd@0: %	Copyright (c) Ian T Nabney (1996-2001)
wolffd@0: 
wolffd@0: unpakstr = [net.type, 'unpak'];
wolffd@0: 
wolffd@0: % Check if we are being passed a masked set of weights
wolffd@0: if (isfield(net, 'mask'))
wolffd@0:    if length(w) ~= size(find(net.mask), 1)
wolffd@0:       error('Weight vector length does not match mask length')
wolffd@0:    end
wolffd@0:    % Do a full pack of all current network weights
wolffd@0:    pakstr = [net.type, 'pak'];
wolffd@0:    fullw = feval(pakstr, net);
wolffd@0:    % Replace current weights with new ones
wolffd@0:    fullw(logical(net.mask)) = w;
wolffd@0:    w = fullw;
wolffd@0: end
wolffd@0: 
wolffd@0: net = feval(unpakstr, net, w);