School of Psychology Colloquium

Presentation by Dr. Jean Philippe Thivierge
Monday, 20 January 2020 - 2:30 pm to 4:00 pm
Location
Contact information
Contact person: 
Patricia Brosseau-Liard
Email: 
psychair@uottawa.ca
Registration
Registration required: 
Yes
Registration details: 

Please register here for the event: https://forms.gle/2ivvjuZ55JHsUXVV9

Cost to attend: 
Free of charge
Event organizer: 
Event language: 

Our second colloquium for 2019-2020 will take place on Monday, January 20th at 2:30pm in FSS 5028.

Our speaker is Dr. Jean-Philippe Thivierge, professor in psychology, University of Ottawa (title and abstract below).

Coffee will be served; bring your mug!

Towards a brain-inspired theory of cognition using recurrent networks

In recent years, advances in artificial intelligence (AI) have opened the possibility of capturing the complex relationship between brain activity, thoughts, and behavior. In this talk, I will argue that novel approaches in this field fail to capture key aspects of learning. These shortcomings require more than minor changes to existing theories, as they represent fundamental limitations to a broad spectrum of models. I present a novel framework that addresses some, though not all, of these limitations. I show results of a novel "reservoir computing" model where information is distributed across a wide network of nodes whose activity reverberates onto itself. Preliminary results based on this framework capture the rapid one-shot learning typical of human behavior and show promising applications to artificial speech processing. This new framework will serve as a springboard for further brain-inspired models of AI that are grounded in cognitive, developmental, and social aspects of behavior.