Model of neural visual system with self-organizing cells (Q1108220)

This paper describes a model of a neural visual system of a higher animal, in which the capability of pattern recognition develops adaptively. To produce the adaptability, we adopted ``self-organizing cells,'' and with them modeled feature-detecting cells which were discovered by \textit{D. Hubel} and \textit{T. Wiesel} [J. Physiol. 160, 106- 154 (1962)] and whose plasticity was found by \textit{C. Blakemore} and \textit{G. Cooper} [Nature 228, 477-478 (1970)]. Combining the ``self-organizing cells'' and the learning principle of a Perceptron-type system, we constructed a model of the whole visual system. The model is also equipped with an eye movement control mechanism for gazing, which reduces the number of ``self-organizing cells'' required for pattern recognition, thus contributing to their quick self- organization. Computer simulation and an experiment using a hardware simulator showed that ``self-organizing cells'' quickly become sensitive to the features often seen and that the resulting system can classify patterns with a rather small number of feature-detecting cells.