Low level modeling of the development of directionally selective microcircuits in cat striate cortex

The development of a directionally selective structure in cat striate cortex is used as a paradigm for exploring issues of biologically plausible and computationally interesting learning algorithms. A compartmental model of a canonical microcircuit is used to represent functional units of the cortex. Initially, the thalamus projects in a random fashion to a pair of cortical microcircuits. Through the use of a sliding threshold model of long-term potentiation and long-term depression, these projections develop biologically plausible, directionally selective responses to randomly moving visual stimuli. The possible implications of the learning algorithm on self-organization in the developing cortex are discussed