Self-Organizing Neural Networks for Stable Control of Autonomous Behavior in A Changing World
Stephen Grossberg · North-Holland mathematical library · 1993
This chapter discusses how neural models function autonomously in a stable fashion despite unexpected changes in their environments. The content of these models consists of a small set of equations that describe processes such as activation of short term memory (STM) traces, associative learning by adaptive weights or long term memory (LTM) traces, and slow habituative gating or medium term memory (MTM) by chemical modulators and transmitters. Two of the core neural network models are often called the additive model and the shunting model. These models explain how STM and LTM traces interact during network processes of activation, associative learning, and recall. One of the most important contributions of neural network models has been to show how behavioral properties can arise as emergent properties because of network interactions. The chapter presents the adaptive resonance theory (ART) and ART systems learn prototypes. These prototypes can be used to encode individual exemplars.