Daniel@0: Daniel@0:
Daniel@0:Daniel@0: w = mlppak(net) Daniel@0:Daniel@0: Daniel@0: Daniel@0:
w = mlppak(net)
takes a network data structure net
and
Daniel@0: combines the component weight matrices bias vectors into a single row
Daniel@0: vector w
. The facility to switch between these two
Daniel@0: representations for the network parameters is useful, for example, in
Daniel@0: training a network by error function minimization, since a single
Daniel@0: vector of parameters can be handled by general-purpose optimization
Daniel@0: routines.
Daniel@0:
Daniel@0: The ordering of the paramters in w
is defined by
Daniel@0:
Daniel@0: Daniel@0: w = [net.w1(:)', net.b1, net.w2(:)', net.b2]; Daniel@0:Daniel@0: Daniel@0: where
w1
is the first-layer weight matrix, b1
is the
Daniel@0: first-layer bias vector, w2
is the second-layer weight matrix,
Daniel@0: and b2
is the second-layer bias vector.
Daniel@0:
Daniel@0: mlp
, mlpunpak
, mlpfwd
, mlperr
, mlpbkp
, mlpgrad
Copyright (c) Ian T Nabney (1996-9) Daniel@0: Daniel@0: Daniel@0: Daniel@0: