Daniel@0: Daniel@0:
Daniel@0:Daniel@0: net = mlpinit(net, prior) Daniel@0:Daniel@0: Daniel@0: Daniel@0:
net = mlpinit(net, prior)
takes a 2-layer feedforward network
Daniel@0: net
and sets the weights and biases by sampling from a Gaussian
Daniel@0: distribution. If prior
is a scalar, then all of the parameters
Daniel@0: (weights and biases) are sampled from a single isotropic Gaussian with
Daniel@0: inverse variance equal to prior
. If prior
is a data
Daniel@0: structure of the kind generated by mlpprior
, then the parameters
Daniel@0: are sampled from multiple Gaussians according to their groupings
Daniel@0: (defined by the index
field) with corresponding variances
Daniel@0: (defined by the alpha
field).
Daniel@0:
Daniel@0:
mlp
, mlpprior
, mlppak
, mlpunpak
Copyright (c) Ian T Nabney (1996-9) Daniel@0: Daniel@0: Daniel@0: Daniel@0: