Adaptive Resonance as a Neural Basis for Conceptual Semantics

Abstract

Experimental data on the interaction between perception and conceptual processing in the domain of motion was modeled by using an adaptive resonance circuit. It was shown that the same mechanisms that are responsible for stable category learning also support perceptual simulation of conceptual knowledge. The model is implemented as a dynamical system, which is able to switch between perception mode, where the network dynamics is driven primarily by the sensory input, and the simulation mode, where the network dynamics is driven by the conceptual input from the internal sources. Computer simulations showed that the differences in reaction times in behavioral studies arise from the match (or mismatch) between the sensory input and top-down learned expectations when the network switches between perception and the simulation mode.


Back to Saturday Papers