The Role of Behavioral Timescale Synaptic Plasticity for Memory Storage in Neural Networks

Abstract

Episodic memory depends crucially on the capacity of neuronal circuits to store information in a one-shot fashion about events that unfold over a time-scale of seconds. Standard Hebbian plasticity rules, such as STDP that require repeated pairing of pre- and post-synaptic activation, are inadequate as physiological mechanisms underlying this type of rapid learning. Contrary to this, Behavioral Timescale Synaptic Plasticity (BTSP), a newly discovered form of plasticity in the hippocampus, operates on a timescale of seconds. This mechanism induces long-lasting synaptic changes after a single experience, driven by dendritic plateau potentials, making it ideally suited for encoding episodic memories. After just one trial, BTSP’s ability to rapidly form place fields in CA1 neurons underscores its critical role in memory formation. This thesis investigates the role of BTSP in memory storage within the hippocampal network. We derive a simplified BTSP model that lends itself to rigorous mathematical analysis, extending this framework to recurrent networks such as the CA3 region of the hippocampus to explore its memory storage properties. Through a detailed examination of recall dynamics, our results demonstrate that BTSP facilitates the encoding and retrieval of a large number of memories, with variability enhancing both storage and recall. Additionally, we explore the non-Hebbian aspect of BTSP, showing that it supports homogeneous representations in CA3. Consequently, we conclude that BTSP is a viable candidate mechanism underlying episodic memory.

Date of Award	11 Nov 2024
Original language	English
Supervisor	Alexander Charles Roxin (Director)