Nodal Chip for Parallel Neural Signal Processing

This work describes a neural network chip which allows combining massive parallel interneural communication with sophisticated intraneural computation and can be used for construction of neural network systems. The chip enables the emulation of different types of neurons including biologically inspired models based on activity pulses, learnable synaptic weights and delays, variable neuron gain, calculation of membrane potential, and static and dynamic thresholding. The chip makes it possible to build neural grid arrays where each node contains one chip and possesses global communication capability despite locally interconnected chips.