Background

Neurons are the basic cells that build the nervous systems of all kinds of animals. If silicon neurons, or neuromorphs, can be endowed with certain essential life-like characteristics, networks of them could be constructed to emulate, or even model the functions of neuronal networks existing in living nervous systems. To this end, we have developed VLSI neuromorphs comprising spatially extensive dendritic trees and spike generating somas. When incorporated into a network connected by a virtual wire system (see Hardware), a neuromorph's pulsatile output can be transmitted over multiple pathways, each of which imparts a programmable delay to the spikes traveling on it.

Our primary research interest is in developing analog VLSI circuits and systems that mimic the nonlinear and spatiotemporal sensory processing capabilities inherent in dendritic trees. To this end we have developed a functional network of neuromorphs and discussed many of its properties (see Publications). 

Currently, we are developing new chips that contain more neuromorphs, and now we have new prototype virtual wire systems that are portable and communicate through a standard interface. These systems are being distributed to interested parties for possible applications and research in areas that require neural networks.