Monthly
288 pp. per issue
6 x 9, illustrated
ISSN
0899-7667
E-ISSN
1530-888X
2014 Impact factor:
2.21

Neural Computation

Spring 1990, Vol. 2, No. 1, Pages 58-70
(doi: 10.1162/neco.1990.2.1.58)
© 1990 Massachusetts Institute of Technology
A Complementarity Mechanism for Enhanced Pattern Processing
Article PDF (622.59 KB)
Abstract

The parallel ON- and OFF-center signals flowing from retina to brain suggest the operation of a complementarity mechanism. This paper shows what such a mechanism can do in higher-level visual processing. In the proposed mechanism, inhibition and excitation, both feedforward, coequally compete within each hierarchical level to discriminate patterns. A computer model tests complementarity in the context of an adaptive, self-regulating system. Three other mechanisms (gain control, cooperativity, and adaptive error control) are included in the model but are described only briefly. Results from simulations show that complementarity markedly improves both speed and accuracy in pattern learning and recognition. This mechanism may serve not only vision but other types of processing in the brain as well.