Model-based analysis of time-dependent consolidation

Foldes, Tamas ORCID: https://orcid.org/0000-0002-0623-9149 2023. Model-based analysis of time-dependent consolidation. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF - Accepted Post-Print Version Download (10MB) | Preview
	PDF (Cardiff University Electronic Publication Form) - Supplemental Material Restricted to Repository staff only Download (101kB)

Abstract

Reactivation and replay in both biological and artificial agents offer significant computational advantages. Research has demonstrated that these processes can lead to accelerated learning, reduced forgetting, and the reorganization or augmentation of experiences, further supporting planning and generalization. These reactivations can transpire during both online and offline periods. Specifically, online reactivation pertains to the immediate reactivation of neural activity patterns during wakefulness. In contrast, offline reactivation takes place during rest or sleep intervals, when the brain is disengaged from external tasks. In these moments, the brain revisits and reactivates neural activity patterns initially established during wakeful states. A significant gap exists in our understanding however of how the brain determines which information to reactivate during its limited offline periods. Specifically, the conditions under which offline reactivation contributes to adaptive generalization remain ambiguous, especially as recent reviews have highlighted the limitations of the extent of offline benefits (Cordi & Rasch, 2021; Lerner & Gluck, 2019a). This uncertainty is not trivial; given the pivotal role offline periods play in memory consolidation, comprehending the mechanisms underlying adaptive consolidation and generalization is paramount. Such understanding can offer insights into optimizing learning strategies and determining the best times to take breaks. Furthermore, it can guide the development of artificial agents designed for continuous learning and might on day serve as reliable assistants in our daily activities. Additionally, a deeper grasp of healthy memory consolidation processes can pave the way for identifying early indicators of consolidation breakdowns, whether due to aging, mental health conditions, or other factors. This knowledge could also usher in innovative applications, such as facilitating learning or unlearning during rest periods, where memory might be more malleable than during wake. The central aim of this dissertation is to investigate the merit of a model-based analysis of offline reactivation-dependent consolidation following episodic learning. This exploration seeks to understand the potential mechanistic contributions of hippocampal offline reactivations to the generalization observed in animals when engaged in episodic and serial learning tasks. Another pivotal objective is to identify moderating variables that can account for why extended post-learning retention intervals, which encompass “offline reactivations,” sometimes result in noticeable generalization benefits, while in other instances they either have no effect or even lead to decreased performance upon delayed retrieval. Augmenting this empirical research with a quantitative meta-analysis can further validate claims regarding these moderator variables. With a deeper understanding of these conditions, it is anticipated that advancements in pattern analysis of generalization, combined with neuroimaging techniques like fMRI, can be employed to pinpoint not only the “loci” of generalization but also the specific nature and evolution of generalization over time. This dissertation, at its core, encapsulates my journey in learning diverse methodologies and approaches with the overarching aim of producing research that is both reproducible and replicable. The significance of this work extends beyond its immediate findings, offering broader implications for the field at large. The cognitive modeling of ostensibly simple tasks, such as transitive inference, holds promise. It not only facilitates precise communication through mathematical paradigms among researchers from disparate disciplines but also engenders interdisciplinary adversarial collaborations. Such collaborations can catalyze the design of experiments situated at the intersection of contending formal theories, thereby fostering incremental advancements in the field. The adoption of online experimentation, particularly within the domain of time and sleep-dependent consolidation, represents a relatively nascent approach. Our endeavors in this realm are anticipated to pave the way for future studies, characterized by enhanced statistical robustness. In tandem with this, our concise meta-analysis of the extant datasets serves as a precursor to more expansive meta-analytical endeavors, poised to elucidate the moderators of generalization and provide direction for subsequent research. Furthermore, this dissertation endeavors to illuminate an efficacious methodology for examining representational shifts over time. This is achieved through a within-subject multi-session design that combines remote learning conditions with in-scanner retrieval, employing a localizer task to scrutinize the representational geometry underpinning inference. While each of these methodological innovations might not be unprecedented in isolation, their confluence within this research area is rare. Such a synthesis holds the potential to augment the current state of sleep and memory research. However, it’s important to note certain limitations that might influence the interpretation or generalizability of the findings. The scope of this dissertation is defined by its concentrated focus on theories of generalization discussed in Chapter 1, particularly emphasizing reactivation as the primary underlying mechanism. In terms of tasks, the research is primarily centered on the phenomena of associative inference, with a specific emphasis on transitive inference throughout. Additionally, while the main emphasis is on offline consolidation, this research does not include any direct physiological measures of reactivation during the offline period. The analysis is anchored solely in recall performance, with only minimal measures following immediate recall. Regrettably, none of the experiments undertaken involve measuring immediate post-learning rest or sleep physiology, which could provide deeper insights into the process of offline consolidation. For this research, a multifaceted methodological approach was adopted to delve into the intricacies of reactivation-dependent generalization. A vector-based memory models were crafted using both MATLAB and Python, facilitating the simulation of specific experiments. These simulations were useful in shedding light on reactivationdependent generalization, resonating with the overarching goals of this dissertation. On the behavioral analysis front, the methodology predominantly hinges on logistic mixed-models and employs remote web-based experimentation to dissect timedependent consolidation. Furthermore, univariate random-effects models have been employed for a meta-analysis of the transitive inference findings, as well as a metaregression analysis of moderator variables. To enhance the depth of the research, a follow-up experiment was analyzed using model-based representational similarity analysis, complemented by fMRI data. The dissertation is structured to provide a rigorous exploration of the topic at hand. Chapter 1 commences with a theoretical exposition, delineating both the classical and more recent paradigms that inform our understanding of human generalization. As the discourse advances, attention is directed towards cognitive models of generalization, with an emphasis on those that can produce offline generalization phenomena. In Chapter 2, a methodical comparative analysis is undertaken, juxtaposing two salient cognitive models: REMERGE and MINERVA2. This chapter underscores the efficacy of model-based methodologies in the study of generalization, positing MINERVA2 as an exemplar baseline model. Its value lies in its capacity to account for an array of time-dependent findings with a parsimonious set of parameters. Chapter 3 presents a triad of empirical investigations centered on the details of time and sleep-dependent generalization as they manifest following the transitive inference task. Of these, one is a reanalysis of an extant dataset, while the subsequent two collected as part of this dissertation, adopt divergent design paradigms: the former adhering to the conventional between-subject design and the latter a novel withinsubject approach. Transitioning to Chapter 4, the narrative engages in a reevaluation of two secondary datasets published in the realm of transitive inference. This chapter ends with a meta-analytical synthesis, amalgamating insights from our three primary experiments, the reinterpreted datasets, and additional published research, thereby facilitating a comprehensive examination of time and sleep-dependent effects. Concluding the dissertation, Chapter 5 introduces the final empirical endeavor, which harnesses neuroimaging techniques to probe the representational geometry underpinning successful inference at delayed test, building upon the previously piloted within-subject study of transitive inference.

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Cardiff University Brain Research Imaging Centre (CUBRIC) Psychology
Funders:	ERC Consolidator Grant 681607 'SolutionSleep'
Date of First Compliant Deposit:	24 April 2024
Last Modified:	25 Apr 2024 01:05
URI:	https://orca.cardiff.ac.uk/id/eprint/168270

Actions (repository staff only)

Edit Item