Date of Award
10-23-2024
Degree Name
Doctor of Philosophy (PhD)
Degree Type
Dissertation
Abstract
A central goal in neuroscience is to understand the relationship between the structure and activity of the brain and the resulting experience of thought and behavior that the brain underpins. Within this goal is the specific study of working memory: the process of storing, managing, and integrating incoming sensory information in the context of ongoing tasks. The process of working memory is foundational to higher cognition and reasoning, determining how we comprehend and interact with the world around us. The neural circuits responsible for working memory are seemingly able to store and process afferent stimuli with other recent or relevant information while concurrently managing limited resources. The question then arises as to how these networks implement these complex and key functions. This dissertation considers how neural circuits could embed functional mnemonic principles within their structure and dynamics, particularly under resource constraints. Using a dynamical systems and optimization framework derived from engineering theory, we synthesize network dynamics that outline and examine central functional principles of encoding and maintaining information in a finite network, and formulate corresponding mathematical objectives. Based on phenomenological concepts of memory, we propose that these systems must specifically balance i) encoding stimuli accurately and efficiently; ii) preventing the corruption of previous representations with new encodings; and iii) preserving these representations over time for potential recall at an unpredictable time. We show that networks optimized from the above principles give rise to architecture and dynamics that manifest features that are comparable to those observed in biology, including gain modulation, lateral inhibition and heterogeneity of time-scales. Critically, these features are linked, through the models, to specific mathematical functional objectives. Furthermore, the models provide links to popular theories of working memory function, including the role of attractors in the network dynamics for memory encoding and storage. In total, the dissertation provides new insights, borne from engineering theory and methods, into how distributed networks can embed high-level memory capability.
Language
English (en)
Chair
ShiNung Ching
Committee Members
Gaia Tavoni; Jr-Shin Li; Lawrence Snyder; Shen Zeng