wolffd@0: wolffd@0: wolffd@0: wolffd@0: Netlab Reference Manual gp wolffd@0: wolffd@0: wolffd@0: wolffd@0:

gp wolffd@0:

wolffd@0:

wolffd@0: Purpose wolffd@0:

wolffd@0: Create a Gaussian Process. wolffd@0: wolffd@0:

wolffd@0: Synopsis wolffd@0:

wolffd@0:

wolffd@0: net = gp(nin, covarfn)
wolffd@0: net = gp(nin, covarfn, prior)
wolffd@0:

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

wolffd@0: Description wolffd@0:

wolffd@0: wolffd@0:

net = gp(nin, covarfn) takes the number of inputs nin wolffd@0: for a Gaussian Process model with a single output, together wolffd@0: with a string covarfn which specifies the type of the covariance function, wolffd@0: and returns a data structure net. The parameters are set to zero. wolffd@0: wolffd@0:

The fields in net are wolffd@0:

wolffd@0:   type = 'gp'
wolffd@0:   nin = number of inputs
wolffd@0:   nout = number of outputs: always 1
wolffd@0:   nwts = total number of weights and covariance function parameters
wolffd@0:   bias = logarithm of constant offset in covariance function
wolffd@0:   noise = logarithm of output noise variance
wolffd@0:   inweights = logarithm of inverse length scale for each input 
wolffd@0:   covarfn = string describing the covariance function:
wolffd@0:       'sqexp'
wolffd@0:       'ratquad'
wolffd@0:   fpar = covariance function specific parameters (1 for squared exponential,
wolffd@0:    2 for rational quadratic)
wolffd@0:   trin = training input data (initially empty)
wolffd@0:   trtargets = training target data (initially empty)
wolffd@0:

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

net = gp(nin, covarfn, prior) sets a Gaussian prior on the wolffd@0: parameters of the model. prior must contain the fields wolffd@0: pr_mean and pr_variance. If pr_mean is a scalar, wolffd@0: then the Gaussian is assumed to be isotropic and the additional fields wolffd@0: net.pr_mean and pr_variance are set. Otherwise, wolffd@0: the Gaussian prior has a mean wolffd@0: defined by a column vector of parameters prior.pr_mean and wolffd@0: covariance defined by a column vector of parameters prior.pr_variance. wolffd@0: Each element of prmean corresponds to a separate group of parameters, which wolffd@0: need not be mutually exclusive. The membership of the groups is defined wolffd@0: by the matrix prior.index in which the columns correspond to the elements of wolffd@0: prmean. Each column has one element for each weight in the matrix, wolffd@0: in the order defined by the function gppak, and each element wolffd@0: is 1 or 0 according to whether the parameter is a member of the wolffd@0: corresponding group or not. The additional field net.index is set wolffd@0: in this case. wolffd@0: wolffd@0:

wolffd@0: See Also wolffd@0:

wolffd@0: gppak, gpunpak, gpfwd, gperr, gpcovar, gpgrad

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

wolffd@0: