Please note: This PhD defence will take place in E7 6443 and online.
Nicole Sandra-Yaffa Dumont, PhD candidate
David R. Cheriton School of Computer Science
Supervisors: Professors Chris Eliasmith, Jeff Orchard
Understanding how the mammalian brain represents and manipulates spatial variables is fundamental to deciphering the mechanisms of navigation and related cognitive functions. This thesis investigates these mechanisms through the lens of Vector Symbolic Algebras (VSAs), specifically focusing on Spatial Semantic Pointers (SSPs) as a biologically plausible framework for neural encoding. Leveraging SSPs, this work develops spiking neural models of path integration, mapping, and reinforcement learning (RL), offering insights into spatial cognition and its broader applications.
The thesis presents key contributions, including a novel SSP-based path integration model, SSP-PI, that integrates continuous spatial variables via velocity-controlled oscillators with attractor dynamics. It also introduces SSP-SLAM, the first biologically inspired spiking semantic SLAM system, capable of constructing cognitive maps that bind spatial and semantic features. Building on this foundation, spiking RL models are proposed, demonstrating how SSP embeddings enable effective representations of successor features, reward distributions, and stochastic policies. Finally, the applicability of SSPs and VSAs is extended to deep RL, showcasing their utility in tasks requiring compositional representations.
These findings underscore the potential of SSPs as a unifying framework for understanding spatial and cognitive processes in the brain while advancing biologically inspired approaches to navigation and learning in artificial systems. This work bridges theoretical neuroscience and artificial intelligence, laying the groundwork for future explorations of shared principles across spatial and abstract cognition.
To attend this PhD defence in person, please go to E7 6443. You can also attend virtually on Zoom.