Daniel@0: Daniel@0:
Daniel@0:Daniel@0: [net] = evidence(net, x, t) Daniel@0: [net, gamma, logev] = evidence(net, x, t, num) Daniel@0:Daniel@0: Daniel@0: Daniel@0:
[net] = evidence(net, x, t)
re-estimates the
Daniel@0: hyperparameters alpha
and beta
by applying Bayesian
Daniel@0: re-estimation formulae for num
iterations. The hyperparameter
Daniel@0: alpha
can be a simple scalar associated with an isotropic prior
Daniel@0: on the weights, or can be a vector in which each component is
Daniel@0: associated with a group of weights as defined by the index
Daniel@0: matrix in the net
data structure. These more complex priors can
Daniel@0: be set up for an MLP using mlpprior
. Initial values for the iterative
Daniel@0: re-estimation are taken from the network data structure net
Daniel@0: passed as an input argument, while the return argument net
Daniel@0: contains the re-estimated values.
Daniel@0:
Daniel@0: [net, gamma, logev] = evidence(net, x, t, num)
allows the re-estimation
Daniel@0: formula to be applied for num
cycles in which the re-estimated
Daniel@0: values for the hyperparameters from each cycle are used to re-evaluate
Daniel@0: the Hessian matrix for the next cycle. The return value gamma
is
Daniel@0: the number of well-determined parameters and logev
is the log
Daniel@0: of the evidence.
Daniel@0:
Daniel@0:
mlpprior
, netgrad
, nethess
, demev1
, demard
Copyright (c) Ian T Nabney (1996-9) Daniel@0: Daniel@0: Daniel@0: Daniel@0: