Neural Pattern Recognition on Multichannel Input Representation

This article presents a new neural pattern recognition architecture on multichannel data representation. The architecture emploies generalized ART modules as building blocks to construct a supervised learning system generating recognition codes on channels dynamically selected in context using serial and parallel match trackings led by inter-ART vigilance signals.