Over the last decades, as the slowdown of Moore’s Law becomes increasingly imminent, various solutions beyond the Turing/von-Neumann computation paradigms have emerged. One of them takes inspiration from the brain—neuromorphic computing—which aims to mimic the collective behavior of biological neurons, synapses, axons, dendrites, etc. The field has grown into a proven, feasible approach but still has remaining challenges such as achieving the energy efficiency of the brain. The key ingredients to develop an energy-efficient neuromorphic computer include new material and functional platforms (where quantum materials have recently become prominent candidates), devices and connectivity types, systems architectures, and mathematical algorithms. It is a timely moment to collect the community’s thoughts on the state-of-the-art of neuromorphic computing and discuss the key advances and difficult challenges to address.
Topics covered include, but are not limited to:
- New, functional, quantum materials beyond CMOS: correlated oxides and other systems, organics, hybrid organic/inorganics perovskites, etc.
- Physical phenomena behind the computation process: memristors, resistive switching, superconducting junctions, magnetic dynamics, optical implementations, etc.,
- Encoding protocols: spiking, analog, reservoir, etc.,
- Network architecture and connectivity implementations in two- and three-dimensions
- Computational and energetic network efficiency
Ivan Schuller, University of California, San Diego
Alex Frano, University of California, San Diego
Jian Shen, Fudan University
Axel Hoffmann, University of Illinois Urbana-Champaign
Robert Dynes, University of California, San Diego
Abu Sebastian, IBM Research – Zurich
Catherine Schuman, Oak Ridge National Laboratory
Anirban Bandyopadhyay, National Institute for Materials Science
Beatriz Noheda, University of Groningen