Liu, Yunzhe;
(2020)
Neural replay in representation, learning and planning.
Doctoral thesis (Ph.D), UCL (University College London).
Preview |
Text
thesis_correction2.pdf - Accepted Version Download (14MB) | Preview |
Abstract
Spontaneous neural activity is rarely the subject of investigation in cognitive neuroscience. This may be due to a dominant metaphor of cognition as the information processing unit, whereas internally generated thoughts are often considered as noise. Adopting a reinforcement learning (RL) framework, I consider cognition in terms of an agent trying to attain its internal goals. This framework motivated me to address in my thesis the role of spontaneous neural activity in human cognition. First, I developed a general method, called temporal delayed linear modelling (TDLM), to enable me to analyse this spontaneous activity. TDLM can be thought of as a domain general sequence detection method. It combines nonlinear classification and linear temporal modelling. This enables testing for statistical regularities in sequences of neural representations of a decoded state space. Although developed for use with human non- invasive neuroimaging data, the method can be extended to analyse rodent electrophysiological recordings. Next, I applied TDLM to study spontaneous neural activity during rest in humans. As in rodents, I found that spontaneously generated neural events tended to occur in structured sequences. These sequences are accelerated in time compared to those that related to actual experience (30 -50 ms state-to-state time lag). These sequences, termed replay, reverse their direction after reward receipt. Notably, this human replay is not a recapitulation of prior experience, but follows sequence implied by a learnt abstract structural knowledge, suggesting a factorized representation of structure and sensory information. Finally, I test the role of neural replay in model-based learning and planning in humans. Following reward receipt, I found significant backward replay of non-local experience with a 160 ms lag. This replay prioritises and facilitates the learning of action values. In a separate sequential planning task, I show these neural sequences go forward in direction, depicting the trajectory subjects about to take. The research presented in this thesis reveals a rich role of spontaneous neural activity in supporting internal computations that underpin planning and inference in human cognition.
| Type: | Thesis (Doctoral) |
|---|---|
| Qualification: | Ph.D |
| Title: | Neural replay in representation, learning and planning |
| Event: | UCL (University College London) |
| Open access status: | An open access version is available from UCL Discovery |
| Language: | English |
| Additional information: | Copyright © The Author 2021. Original content in this thesis is licensed under the terms of the Creative Commons Attribution 4.0 International (CC BY 4.0) Licence (https://creativecommons.org/licenses/by/4.0/). Any third-party copyright material present remains the property of its respective owner(s) and is licensed under its existing terms. Access may initially be restricted at the author’s request. |
| UCL classification: | UCL UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences > Faculty of Brain Sciences UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences > Faculty of Brain Sciences > UCL Queen Square Institute of Neurology UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences > Faculty of Brain Sciences > UCL Queen Square Institute of Neurology > Imaging Neuroscience |
| URI: | https://discovery.ucl.ac.uk/id/eprint/10118391 |
Archive Staff Only
 |
View Item |
