GTM

Trainable mapping fitting a Generative Topographic Mapping by EM

    [W,L] = GTM (A,K,M,MAPTYPE,REG,EPS,MAXITER)

Input
A	Dataset or double matrix
K	Vector containing number of nodes per dimension (default: [5 5], 2D map)
M	Vector containing number of basis functions per dimension (default: [10 10])
MAPTYPE	Map onto mean of posterior ('mean', default) or mode ('mode')
REG	Regularisation (default: 0)
EPS	Change in likelihood to stop training (default: 1e-5)
MAXITER	Maximum number of iterations (default: inf)

Output
W	GTM mapping
L	Likelihood

Description

Trains a Generative Topographic Mapping of any dimension, using the EM algorithm.

Reference(s)

Bishop, C.M., Svensen, M. and Williams, C.K.I., "GTM: The Generative Topographic Mapping", Neural Computation 10(1):215-234, 1998.

See also

plotgtm, som, prplotsom,