The Ability of Separability in a Layered Random Network

The cortex of the vertebrate cerebellum has a simple and extremely fine struture. The system is organized with the mossy fibers, granule cells, Golgi cells, basket cells, and Purkinje cells. Functionally, the cortex of cerebellum works as a pattern-recognition data processing system. Moreover, this system (network) has two eminent abilities which makes the overlap of two input patterns reduce and makes the learning enhance. The present paper gives a fundamental and theoretical analysis of the ability of the former funtion and some results of computer-simulated experiment.