Pattern recognition in sequences of neuronal spike intervals; a new nonparametric approach and its implications
Marczynski, T.J.
Materia Medica Polona. Polish Journal of Medicine and Pharmacy 8(2): 97-107
1976
ISSN/ISBN: 0025-5246 PMID: 790034 Document Number: 108982
New computer programs were described for the analysis of temporal patterns in long sequences of extracellularly recorded neuronal spike intervals. These pattern recognition techniques were based on a non-parametric principle: the computer first measures the duration of each interval, then compares the sequential pairs and makes inequality statements about them. Subsequently, these statements were arranged in various transition-frequency matrices from which the probability of each pattern was calculated. The techniques also allowed the computation of the theoretical, i.e., chance, probabilities of each encountered pattern, assuming that the sequential arrangement of intervals is totally random and/or independent. The results lend themselves to the interpretation by classic information theory concepts derived from Shannon's formula for uncertainty or entropy in a communication system. As an example patterns were described in the output of 2 neurons from the caudate nucleus. Two working hypotheses were proposed: one concerning the basic mechanism of reinforcement and learning, and the other explaining the nature of the recuperative role of slow-wave sleep.