Pattern-Generator-Driven Development in Self-Organizing Models

Abstract

ABSTRACTSelf-organizing models develop realistic cortical structures whengiven approximations of thevisual environment as input. Recently it has been proposed that internallygenerated input pat-terns, such as those found in the developing retina and in PGO wavesduring REM sleep, mayhave the same effect. Internal pattern generators would constitutean efficient way to specify,develop, and maintain functionally appropriate perceptual organization.They may help expresscomplex structures from minimal genetic information, and retain thisgenetic structure within ahighly plastic system. Simulations with the RF-LISSOM orientationmap model indicate thatsuch preorganization is possible, providing a computational frameworkfor examining how ge-netic influences interact with visual experience.