ABSTRACT
This chapter discusses an algorithm developed by Crutchfield and Young to produce symbolic descriptions of the behaviour of a simple neuralnetwork model of the cerebral cortex. It explores some of the conditions under which the computational power of the symbolic interpretation of the network’s behaviour is maximized, relating this computational power to the computational demands of cognitive processes. Symbolic models of cognition are based on collections of rules which determine how a model responds as it encounters stimuli. Neural networks appear to offer an alternative approach which might avoid combinatorial explosion when dealing with complex categorizations. Neural networks consist of a large number of neuron-like elements connected together by links through which the activity of one unit influences the activities of those it is connected to. Entities are represented in neural networks as patterns of activation in these units.
