wolffd@0: wolffd@0: wolffd@0: wolffd@0: Netlab Reference Manual evidence wolffd@0: wolffd@0: wolffd@0: wolffd@0:

evidence wolffd@0:

wolffd@0:

wolffd@0: Purpose wolffd@0:

wolffd@0: Re-estimate hyperparameters using evidence approximation. wolffd@0: wolffd@0:

wolffd@0: Synopsis wolffd@0:

wolffd@0:

wolffd@0: [net] = evidence(net, x, t)
wolffd@0: [net, gamma, logev] = evidence(net, x, t, num)
wolffd@0:

wolffd@0: wolffd@0: wolffd@0:

wolffd@0: Description wolffd@0:

wolffd@0: [net] = evidence(net, x, t) re-estimates the wolffd@0: hyperparameters alpha and beta by applying Bayesian wolffd@0: re-estimation formulae for num iterations. The hyperparameter wolffd@0: alpha can be a simple scalar associated with an isotropic prior wolffd@0: on the weights, or can be a vector in which each component is wolffd@0: associated with a group of weights as defined by the index wolffd@0: matrix in the net data structure. These more complex priors can wolffd@0: be set up for an MLP using mlpprior. Initial values for the iterative wolffd@0: re-estimation are taken from the network data structure net wolffd@0: passed as an input argument, while the return argument net wolffd@0: contains the re-estimated values. wolffd@0: wolffd@0:

[net, gamma, logev] = evidence(net, x, t, num) allows the re-estimation wolffd@0: formula to be applied for num cycles in which the re-estimated wolffd@0: values for the hyperparameters from each cycle are used to re-evaluate wolffd@0: the Hessian matrix for the next cycle. The return value gamma is wolffd@0: the number of well-determined parameters and logev is the log wolffd@0: of the evidence. wolffd@0: wolffd@0:

wolffd@0: See Also wolffd@0:

wolffd@0: mlpprior, netgrad, nethess, demev1, demard

wolffd@0: Pages: wolffd@0: Index wolffd@0:

wolffd@0: