Neural Darwinism and Selective Recognition Automata: How Selection Shapes Perceptual Categories

Abstract

It is becoming increasingly clear that animals generally do not use classical algorithmic strategies to classify objects and events. In this chapter, we summarize an alternative approach, namely, the theory of neuronal group selection (TNGS). The TNGS provides a framework related to Darwinian selection for understanding perceptual systems without a priori assumptions about properties of the stimulus world. We present two models for aspects of visual perception based on synthetic neural modelling, a paradigm we have developed for testing theories about the nervous system by computer simulation. One of these models explores the implications of recent experimental evidence for coherent oscillations in visual cortex. The second is Darwin III, a neurally based simulated automaton capable of autonomous behavior involving categorization and motor acts with an eye and an arm. Automata of this type not only provide new insights into biological mechanisms of categorization, but also indicate how to construct a new class of perception machines with interesting properties.