Plato's Cave: The Boundary Contour System

The Boundary Contour System (BCS) Model

Grossberg's BCS model accounts for collinear illusory contours such as the Kanizsa figure in a dynamic neural network architecture. The principle of operation involves a layer of oriented edge sensitive cells, which respond to the visible edges in the Kanizsa figure, and a higher level cooperative cell layer which receives activation from the oriented cell layer by way of a bipolar receptive field in an orientation specific manner. For example a horizontal cooperative cell receives input only from horizontal oriented cells which are horizontally disposed in the oriented layer, as shown below. This ensures that the cooperative cell will only become active along extended collinear edges. The cooperative cell in turn provides top-down feedback to the corresponding oriented cell in the oriented layer, as shown in the figure. It is this top-down feedback which accounts for the appearence of an illusory contour in a region between two collinear edges as in the Kanizsa figure.

A conjunctive constraint, or functional AND-gate is also incorporated into the model in order to prevent cooperative activation from extending outwards from isolated edge segments.

In the full Boundary Contour System model there are many other layers and interactions that serve various purposes.

Return to argument

Return to Steve Lehar