REPLICATOR NEURAL NETWORKS FOR UNIVERSAL OPTIMAL SOURCE-CODING

Replicator neural networks self-organize by using their inputs as desi red outputs; they internally form a compressed representation for the input data. A theorem shows that a dass of replicator networks can, th rough the minimization of mean squared reconstruction error (for insta nce, by training on raw data examples), carry out optimal data compres sion for arbitrary data vector sources. Data manifolds, a new general model of data sources, are then introduced and a second theorem shows that, in a practically important limiting case, optimal-compression re plicator networks operate by creating an essentially unique natural co ordinate system for the manifold.