wolffd@0: net = gtminit(net, options, data, samptype) wolffd@0: net = gtminit(net, options, data, samptype, lsampsize, rbfsampsize) wolffd@0:wolffd@0: wolffd@0: wolffd@0:
net = gtminit(net, options, data, samptype) takes a GTM net
wolffd@0: and generates a sample of latent data points and sets the centres (and
wolffd@0: widths if appropriate) of
wolffd@0: net.rbfnet.
wolffd@0:
wolffd@0: If the samptype is 'regular', then regular grids of latent
wolffd@0: data points and RBF centres are created. The dimension of the latent data
wolffd@0: space must be
wolffd@0: 1 or 2. For one-dimensional latent space, the lsampsize parameter
wolffd@0: gives the number of latent points and the rbfsampsize parameter
wolffd@0: gives the number of RBF centres. For a two-dimensional latent space,
wolffd@0: these parameters must be vectors of length 2 with the number of points
wolffd@0: in each of the x and y directions to create a rectangular grid. The
wolffd@0: widths of the RBF basis functions are set by a call to rbfsetfw
wolffd@0: passing options(7) as the scaling parameter.
wolffd@0:
wolffd@0:
If the samptype is 'uniform' or 'gaussian' then the
wolffd@0: latent data is found by sampling from a uniform or
wolffd@0: Gaussian distribution correspondingly. The RBF basis function parameters
wolffd@0: are set
wolffd@0: by a call to rbfsetbf with the data parameter
wolffd@0: as dataset and the options vector.
wolffd@0:
wolffd@0:
Finally, the output layer weights of the RBF are initialised by wolffd@0: mapping the mean of the latent variable to the mean of the target variable, wolffd@0: and the L-dimensional latent variale variance to the variance of the wolffd@0: targets along the first L principal components. wolffd@0: wolffd@0:
gtm, gtmem, pca, rbfsetbf, rbfsetfwCopyright (c) Ian T Nabney (1996-9) wolffd@0: wolffd@0: wolffd@0: wolffd@0: