Use this URL to cite or link to this record in EThOS:
Title: Palimpsest working memory
Author: Matthey-De-L'Endroit, Loïc
ISNI:       0000 0004 8499 9420
Awarding Body: UCL (University College London)
Current Institution: University College London (University of London)
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Despite its key role in cognition, the mechanisms underlying working memory remain much debated. It has often been observed that human performance on memory tasks is severely limited, but the two main classes of theories examining these limits leave open many key issues into their underlying sources. More specifically, the question of how multiple stimuli are represented and distinguished in visual working memory is still not well understood. As a first attempt at tackling these issues, we introduce a probabilistic palimpsest model which uses the activity of a single population of neurons to encode several multi-featured items. This population is used in a probabilistic framework to store and recall visual stimuli on a trial-by-trial basis, making it possible to account for many qualitative aspects of existing experimental data. In our setting, the underlying nature of a memory item, and the interference between concurrent stimuli, depend entirely on the characteristics of the population representation. We explore how much can be explained about the patterns of errors observed in human reports purely from the representations being used, without explicitly addressing how recall mechanisms could affect it. We provide analytical and numerical insights into critical issues such as multiplicity and binding. We consider different types of population codes, where information about individual feature values is partially separate from the information about binding that creates single items out of multiple features. We find that a tight balance between these two types of information is required to capture the different types of error seen in human experimental data fully. Our model can also be readily extended to sequentially presented data, making full use of our palimpsest construction. This allows us to study and account for experimental data that have not previously been explored extensively. We argue that our work constitutes an important step towards mechanistic models of visual working memory that provide a more holistic account of human responses based on computational principles.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available