This talk will introduce the Adaptive Resonance Theory (ART) family of memory models.  ART models provide the most accessible entry point into the large array of neural network models pioneered by Stephen Grossberg.  ART models are classifiers built from neurophysiologically plausible neurons. These classifiers are unsupervised not only in the sense that they learn patterns without an external teacher, but also in the sense that input patterns need not be artificially separated.  This talk will cover cooperative-competitive fields (a.k.a. center/surround anatomies), pre- and post-postsynaptic activity gated STM->LTM transfer, and the ART(1) architecture.  It will also touch on later ART models, an adaptive user interface product which uses ART, and related neural models.

Return to the seminar schedule