Cloning cerebellar cortex organization with multi-layer cellular neural network (CNN)

摘要

A multilayer cellular neural network (CNN) architecture that takes the cerebellar cortex as a template is described. Massively parallel systems such as cellular automata (CAs) are well suited to implementing any kind of neural network architecture: fully connected or layered, including or not including recursive branches. Using a simple CA accelerator, neural networks from 1000 to 1,000,000 neurons can easily be implemented.