Bio-inspired ultra lower-power neuromorphic computing engine for embedded systems

Abstract

Neuromorphic computing, which is inspired by the working mechanism of human brain, recently emerges as a hot research area to combat the contradiction between the limited functions of computing systems and the ever increasing variety of applications. In this work, we will introduce our research on a bio-inspired neuromorphic embedded computing engine named Centaur, which aims to achieve an ultra-high power efficiency beyond One-TeraFlops-Per-Watt by adopting the bio-inspired computation model and the advanced memristor technology. The success of Centaur design may promote the embedded system power efficiency three orders of magnitude from the current level while the small footprint and real-time re-configurability of the design allow an easy integration into MPSoCs, enabling many emerging mobile and embedded applications.