A Special Role for Input Codes in Solving the Transverse Patterning Problem

Abstract

Rats require a hippocampus to solve the transverse patterning problem. Here, a hippocampal model also solves this configural learning problem. The problem is hard: A learning paradigm, called progressive learning, is required. It is required by rats, humans, and the model. Second, input patterns within a sequence must be repeated. Such repetition increases the statistical dependence, a surprising observation if you assume statistical dependence is undesirable. Such repetition of the same patterns in a sequence facilitates the formation of local context neuronal firings. These neuronal firings are critical, and we hypothesize that they are analogous to place cells found in behaving animals.