From basic network principles to neural architecture: emergence of spatial-opponent cells.

Abstract

The functional architecture of mammalian visual cortex has been elucidated in impressive detail by experimental work of the past 20-25 years. The origin of many of the salient features of this architecture, however, has remained unexplained. This paper is the first of three (the others will appear in subsequent issues of these Proceedings) that address the origin and organization of feature-analyzing (spatial-opponent and orientation-selective) cells in simple systems governed by biologically plausible development rules. I analyze the progressive maturation of a system composed of a few layers of cells, with connections that develop according to a simple set of rules (including Hebb-type modification). To understand the prenatal origin of orientation-selective cells in certain primates, I consider the case in which there is no external input, with the first layer exhibiting random spontaneous electrical activity. No orientation preference is specified to the system at any stage, and none of the basic developmental rules is specific to visual processing. Here I introduce the theory of "modular self-adaptive networks," of which this system is an example, and explicitly demonstrate the emergence of a layer of spatial-opponent cells. This sets the stage for the emergence, in succeeding layers, of an orientation-selective cell population.