From Brain Modeling to State-of-the-art AI: The story of the LMU

Speaker:  Chris Eliasmith, University of Waterloo

Date: September, 21
Time: 4 pm
Room: LH4114 (Lazaridis Hall, Room 4114)

Abstract:: How much we can learn about building better intelligent systems from looking at the biological brain is a matter of considerable debate. In this talk, I provide a clear, recent example of how understanding biological computation has given rise to a new class of powerful, state-of-the-art
neural networks. This example is the Legendre Memory Unit (LMU), which was derived from a characterization of neural dynamics that capture 'time cells in the brain and has recently been used to set a variety of state-of-the-art results on machine learning benchmarks. I describe the LMU, its derivation, its relationship to neural data, and some of its recent success in AI.

Bio: Professor Chris Eliasmith is the founding Director of the Centre for Theoretical Neuroscience at the University of Waterloo and holds the Canada Research Chair in Theoretical Neuroscience. He has authored two books and over 130 publications in philosophy, psychology, neuroscience, computer science, and engineering venues. He is also a co-founder of Applied Brain Research, Inc. a local deep-tech startup.