SYDE Graduate Seminar - A Biologically Plausible, Dynamics-based Model of Memory Using Neurosymbolic Methods

Presenter

Jakeb Chouinard, MASc candidate in Systems Design Engineering

Abstract

We present a novel model of memory that addresses previous gaps in working memory modelling. Traditionally, models of memory treat memory as an atemporal process; they present a vector or trace of a memory that only changes with item presentations. We instead provide a dynamics-based model of short-term memory using vector projection, and we further characterize its capacity and forgetting mechanisms using temporal decay and neural saturation. We also provide an auto-associative memory that uses bundles of high-dimensional vectors to implicitly associate items to each other, enabling bundle reconstruction from single components. Lastly, we modify the parameters of the short-term memory to create a primacy-biased integrator of task-contextual embeddings. Our approach allows us to implement the model in spiking neurons and validate the model against human behavioural data, showing significant coherence to human data on serial and free recall task paradigms as well as a spatial positional recall task paradigm with minimal model changes between tasks.

Join us on Zoom or in-person in EC4-2031

Meeting ID: 938 8265 5806
Passcode: 111553

This seminar counts towards the graduate student seminar attendance milestone!