Dynamics of Mode-Switching in Neural Networks: How nonlinear systems generate behaviour via a thermodynamic process
Abstract/Contents
- Abstract
- Animals engage in specific motivated actions, such as feeding and drinking, to maintain homeostasis. However, the neural correlates of motivational states--and the mechanisms through which animals transduce internal need into action--remain largely unknown. In this thesis, we study motivational drives by examining what happens when they come into conflict. We begin by developing a simple behavioural task to interrogate the balance of homeostatic needs in mice. While the mouse performs this task, we record activity from a large population of neurons distributed across its brain. We characterize the dynamics of these neurons using dimensionality reduction techniques, with the goal of distilling our dataset into its most informative components. To understand the mechanism underlying behavioural mode-switching, we train a recurrent neural network to capture the salient features of its biological counterpart. We then reverse-engineer this computational model by linearizing in the vicinity of its stable states, demonstrating the existence of two basins of attraction--corresponding, physically, to food-seeking and water-seeking behaviour--joined by a smooth line of intermediate semi-stable states. Finally, we discuss how the system's motion along this continuous manifold can give rise to two robust and distinct actions: cue-driven eating and drinking. In order to do so, we recast our discussion of homeostatic need selection as a thermodynamic process, illustrating this analogy with an out-of-equilibrium Ising model that reproduces the relevant behaviour of our biological system.
Description
| Type of resource | text |
|---|---|
| Date created | May 2020 |
| Date modified | December 5, 2022 |
| Publication date | January 1, 2022 |
Creators/Contributors
| Author | Ticea, Nicole Sabina |
|---|---|
| Degree granting institution | Stanford University, Department of Physics |
| Thesis advisor | Luo, Liqun |
| Thesis advisor | Deisseroth, Karl |
| Thesis advisor | Kachru, Shamit |
Subjects
| Subject | Computational neuroscience |
|---|---|
| Subject | Theoretical neuroscience |
| Subject | Nonlinear dynamical systems |
| Genre | Text |
| Genre | Thesis |
Bibliographic information
Access conditions
- Use and reproduction
- User agrees that, where applicable, content will not be used to identify or to otherwise infringe the privacy or confidentiality rights of individuals. Content distributed via the Stanford Digital Repository may be subject to additional license and use restrictions applied by the depositor.
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Preferred citation
- Preferred citation
- Nicole Sabina Ticea. (2020). Dynamics of Mode-Switching in Neural Networks: How nonlinear systems generate behaviour via a thermodynamic process. Stanford Digital Repository. Available at: https://purl.stanford.edu/bt112tm5286
Collection
Undergraduate Theses, Department of Physics
View other items in this collection in SearchWorksContact information
- Contact
- nicole.ticea5@gmail.com
Also listed in
Loading usage metrics...