Neuromorphic Hardware

Modern electronics have evolved through a series of major developments (e.g., transistors, integrated circuits, microcircuits) that have led to the programmable electronic machines that are ubiquitous today.  Because of limitations in hardware and architecture, these machines are of limited utility in complex, real-world environments.

To address this deficiency, CNES is developing large-scale neuromorphic electronics that can mimic brain-like dynamics and efficiencies based on the computing paradigm developed in our Neural and Cognitive Systems research.

We are currently exploring a hybrid approach (with analog and digital properties) based on CMOS technology with design innovations such as synaptic time multiplexing, floating gate synapses, pulse-processing neurons and synapses and CMOS interconnect fabric with virtual connectivity to emulate the hybrid and distributed architecture features found in the brain. We are also exploring the integration of mixed-signal circuits with advanced nanomaterial structures to enable high connectivity, high component densities and low power.

It is likely that the brain’s efficiency (power  <40 W) is due to computing in such a hybrid fashion with analog and digital properties in a fully distributed architecture.  We envision a path to design and fabricate scalable neuromorphic architectures that will one day mimic mammalian brains in intelligence and efficiency.


Program and Sponsor

Project: Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE)
Sponsor: Defense Advanced Research Projects Agency (DARPA)
Partners: Neurosciences Institute, Boston University, University of California Irvine, Georgia Institute of Technology, George Mason University, University of Michigan, SET Corporation, Portland State University, University of Nevada Reno, Stanford University