Advancing models of the visual system using biologically plausible unsupervised spiking neural networks
<p>Spikes are thought to provide a fundamental unit of computation in the nervous system. The retina is known to use the relative timing of spikes to encode visual input, whereas primary visual cortex (V1) exhibits sparse and irregular spiking activity – but what do these different spiking pat...
| Автор: | |
|---|---|
| Інші автори: | |
| Формат: | Дисертація |
| Мова: | English |
| Опубліковано: |
2023
|
| Предмети: |
| _version_ | 1826313258301128704 |
|---|---|
| author | Taylor, L |
| author2 | King, A |
| author_facet | King, A Taylor, L |
| author_sort | Taylor, L |
| collection | OXFORD |
| description | <p>Spikes are thought to provide a fundamental unit of computation in the nervous system. The retina is known to use the relative timing of spikes to encode visual input, whereas primary visual cortex (V1) exhibits sparse and irregular spiking activity – but what do these different spiking patterns represent about sensory stimuli? To address this question, I set out to model the retina and V1 using a biologically-realistic spiking neural network (SNN), exploring the idea that temporal prediction underlies the sensory transformation of natural inputs.</p>
<p>Firstly, I trained a recurrently-connected SNN of excitatory and inhibitory units to predict the sensory future in natural movies under metabolic-like constraints. This network exhibited V1-like spike statistics, simple and complex cell-like tuning, and - advancing prior studies - key physiological and tuning differences between excitatory and inhibitory neurons.</p>
<p>Secondly, I modified this spiking network to model the retina to explore its role in visual processing. I found the model optimized for efficient prediction to capture retina-like receptive fields and - in contrast to previous studies - various retinal phenomena, such as latency coding, response omissions, and motion-tuning properties. Notably, the temporal prediction model also more accurately predicts retinal ganglion cell responses to natural images and movies across various animal species.</p>
<p>Lastly, I developed a new method to accelerate the simulation and training of SNNs, obtaining a 10-50 times speedup, with performance on a par with the standard training approach on supervised classification benchmarks and for fitting electrophysiological recordings of cortical neurons.</p>
<p>The retina and V1 models lay the foundation for developing normative models of increasing biological realism and link sensory processing to spiking activity, suggesting that temporal prediction is an underlying function of visual processing. This is complemented by a new approach to drastically accelerate computational research using SNNs.</p> |
| first_indexed | 2024-09-25T04:11:53Z |
| format | Thesis |
| id | oxford-uuid:3e6491d1-570e-41ce-b4b9-5b676c801a77 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-09-25T04:11:53Z |
| publishDate | 2023 |
| record_format | dspace |
| spelling | oxford-uuid:3e6491d1-570e-41ce-b4b9-5b676c801a772024-06-25T09:09:00ZAdvancing models of the visual system using biologically plausible unsupervised spiking neural networksThesishttp://purl.org/coar/resource_type/c_db06uuid:3e6491d1-570e-41ce-b4b9-5b676c801a77Machine learningNeuroscienceComputational neuroscienceEnglishHyrax Deposit2023Taylor, LKing, AHarper, NZenke, F<p>Spikes are thought to provide a fundamental unit of computation in the nervous system. The retina is known to use the relative timing of spikes to encode visual input, whereas primary visual cortex (V1) exhibits sparse and irregular spiking activity – but what do these different spiking patterns represent about sensory stimuli? To address this question, I set out to model the retina and V1 using a biologically-realistic spiking neural network (SNN), exploring the idea that temporal prediction underlies the sensory transformation of natural inputs.</p> <p>Firstly, I trained a recurrently-connected SNN of excitatory and inhibitory units to predict the sensory future in natural movies under metabolic-like constraints. This network exhibited V1-like spike statistics, simple and complex cell-like tuning, and - advancing prior studies - key physiological and tuning differences between excitatory and inhibitory neurons.</p> <p>Secondly, I modified this spiking network to model the retina to explore its role in visual processing. I found the model optimized for efficient prediction to capture retina-like receptive fields and - in contrast to previous studies - various retinal phenomena, such as latency coding, response omissions, and motion-tuning properties. Notably, the temporal prediction model also more accurately predicts retinal ganglion cell responses to natural images and movies across various animal species.</p> <p>Lastly, I developed a new method to accelerate the simulation and training of SNNs, obtaining a 10-50 times speedup, with performance on a par with the standard training approach on supervised classification benchmarks and for fitting electrophysiological recordings of cortical neurons.</p> <p>The retina and V1 models lay the foundation for developing normative models of increasing biological realism and link sensory processing to spiking activity, suggesting that temporal prediction is an underlying function of visual processing. This is complemented by a new approach to drastically accelerate computational research using SNNs.</p> |
| spellingShingle | Machine learning Neuroscience Computational neuroscience Taylor, L Advancing models of the visual system using biologically plausible unsupervised spiking neural networks |
| title | Advancing models of the visual system using biologically plausible unsupervised spiking neural networks |
| title_full | Advancing models of the visual system using biologically plausible unsupervised spiking neural networks |
| title_fullStr | Advancing models of the visual system using biologically plausible unsupervised spiking neural networks |
| title_full_unstemmed | Advancing models of the visual system using biologically plausible unsupervised spiking neural networks |
| title_short | Advancing models of the visual system using biologically plausible unsupervised spiking neural networks |
| title_sort | advancing models of the visual system using biologically plausible unsupervised spiking neural networks |
| topic | Machine learning Neuroscience Computational neuroscience |
| work_keys_str_mv | AT taylorl advancingmodelsofthevisualsystemusingbiologicallyplausibleunsupervisedspikingneuralnetworks |