Brain Inspired Computing: the BrainScaleS-2 project & custom ASICs for accelerated analog neuromorphic computing

by Dr Johannes Schemmel (KIP Heidelberg)

Tuesday 13 Apr 2021, 09:00 → 10:00 Europe/Zurich

Zoom details are in the invitation email.

Make sure you have the latest Zoom client app to be able to connect https://cern.zoom.us/ -> Download client

Abstract:

Brain Inspired or Neuromorphic Computing, as a realization of Non-Turing, in-memory, event-based computing, will allow us to overcome the power wall our CPU-centric CMOS technology is facing. But that does not mean that the era of Turing-based computing will come to an end soon, or that Turing-based computing does not have its place in the neuromorphic world. This talk will shortly summarize how the Heidelberg BrainScaleS-2 accelerated analog neuromorphic architecture balances Turing and Non-Turing computing to combine power efficiency with the necessary flexibility and programmability, thereby reducing the resource requirements of AI and extending it by recent insights from neuroscience. These bio-inspired AI technologies may be beneficial for the data challenges the next generation of science instrumentation is facing. Possible applications of the BrainScaleS technology in the areas of edge computing and high-energy physics will be presented.

BrainScaleS - custom ASICs for accelerated analog neuromorphic computing_public.pdf