MSc thesis presentation - Ryan Fayyazi

Date

Name: Ryan Fayyazi

Date: Monday Sept 9th 2024

Time: 10AM

Zoom link: https://ubc.zoom.us/j/68090823333?pwd=orJ2HfbYNJkxSbqOOX7hkMmTzrwwKw.1  

Meeting ID: 680 9082 3333
Passcode: 517501

Supervisor: Frank Wood

 

Thesis Title: Learning in Networks with Communication Delays

 

Abstract: Communication delays are ubiquitous in physically realized neural networks such as biological neural circuits and neuromorphic hardware. These delays have sig- nificant and often disruptive consequences on network dynamics during training and inference. It is therefore essential that communication delays be accounted for both in computational models of biological neural networks, and in large-scale neuromorphic systems. Nonetheless, communication delays have yet to be com- prehensively addressed in either domain. In this thesis, we first show that delays prevent state-of-the-art continuous-time neural networks called Latent Equilibrium (LE) networks from learning even simple tasks despite significant overparameter- ization. We then propose to compensate for communication delays by predicting future signals based on recently received ones. This conceptually straightforward approach, which we call prospective messaging (PM), uses only information local to each synapse or neuron and is flexible in terms of memory and computation requirements. We demonstrate that incorporating PM into delayed LE networks prevents reaction lags, and facilitates successful learning on Fourier synthesis and autoregressive video prediction tasks.