Scaling Up to Large Neuromorphic Systems

Organizer:

Sumedh Risbud

Time:

11:45 am - 12:15 pm (PT), Fri, Feb 12, 2021

Abstract:

Through a discussion in this session, the attendees will learn about benefits and challenges in scaling spiking neural networks to multi-chip neuromorphic systems. Benefits of scaling to large systems are fairly obvious. As for challenges, we will explore trade-offs between: time spent in communication vs computation, efficient allocation of neural vs synaptic resources, on-chip computation vs off-chip computation (as necessitated by bandwidth constraints in I/O), etc. Knowing that such trade-offs exist, is a valuable insight.

Panelists:

• J. Brad Aimone (Sandia National Lab)

• NCL team members (Garrick Orchard, Andreas Wild, Prasad Joshi, Mike Davies)

Pre-requisites/

co-requisites:

None

Recording:

Coming soon

Link to Presentation:

Coming soon

Previous INRC Forum Presentations by the Panelists:

• A Neuromorphic Future for Classic Computing Tasks
  J. Brad Aimone

• Neuromorphic nearest neighbor search on Intel’s Pohoiki Springs
  Garrick Orchard, E. Paxon Frady, Mike Davies