Competitive learning in rate coded and spiking neural networks models with applications to vision and audition
<p>Competitive learning is a common and successful approach used to train unsupervised rate-based neural network models. We apply such a technique in this thesis and produce a rate-coded neural network model of pitch processing which provides insights into the training protocols necessary to d...
| Egile nagusia: | |
|---|---|
| Beste egile batzuk: | |
| Formatua: | Thesis |
| Hizkuntza: | English |
| Argitaratua: |
2019
|
| Gaiak: |
| _version_ | 1826303549646045184 |
|---|---|
| author | Ahmad, N |
| author2 | Stringer, S |
| author_facet | Stringer, S Ahmad, N |
| author_sort | Ahmad, N |
| collection | OXFORD |
| description | <p>Competitive learning is a common and successful approach used to train unsupervised rate-based neural network models. We apply such a technique in this thesis and produce a rate-coded neural network model of pitch processing which provides insights into the training protocols necessary to develop robust pitch representations. However, the extension of reliable unsupervised competitive learning approaches from rate-coded to spiking neural networks has proven challenging, especially when biological plausibility and detail are desired. Transitioning to spiking neural network models is made more difficult by the comparatively high computational cost and complexity of these network models compared to rate-based models.</p>
<p>We describe a transition from rate-based to spiking neural network models and address these outstanding issues. First, we focus on increasing simulation efficiency. We develop a state of the art graphical processing unit (GPU) based spiking neural network simulator which adopts optimisations from central processing unit (CPU) and cluster-based simulators. In benchmarks, we show that our novel simulator is capable of simulation speeds up to an order of magnitude greater than contemporary simulators. This greater simulation speed is intended to enable a higher throughput of spiking neural network simulations and thereby accelerate research. In order to ensure that present and future simulators can be efficiently and effectively compared, we also compile a repository of benchmarks which allow validation of simulator performance and simulated neural network dynamics.</p>
<p>Having developed efficient simulation techniques, we propose a novel excitatory plasticity rule which, when used to train spiking neural networks in conjunction with an existing inhibitory plasticity rule, produces reliable competitive learning. These rules are both unsupervised, use local only information, and depend upon spike-timing in order to compute synaptic weight updates. This learning rule framework is used to produce a model of V1 simple cell receptive field development and shows qualitatively similar behaviour to traditional sparse coding models. During the development of these models we come across a number of challenges, especially in the interactions of inhibitory and excitatory neurons. We pose solutions to these challenges and thereafter suggest some future research questions and avenues for exploration.</p> |
| first_indexed | 2024-03-07T06:04:23Z |
| format | Thesis |
| id | oxford-uuid:ed5434ec-cfe4-43b9-93e5-e07d5617b042 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T06:04:23Z |
| publishDate | 2019 |
| record_format | dspace |
| spelling | oxford-uuid:ed5434ec-cfe4-43b9-93e5-e07d5617b0422022-03-27T11:24:10ZCompetitive learning in rate coded and spiking neural networks models with applications to vision and auditionThesishttp://purl.org/coar/resource_type/c_db06uuid:ed5434ec-cfe4-43b9-93e5-e07d5617b042Computational NeuroscienceEnglishHyrax Deposit2019Ahmad, NStringer, SWalker, K<p>Competitive learning is a common and successful approach used to train unsupervised rate-based neural network models. We apply such a technique in this thesis and produce a rate-coded neural network model of pitch processing which provides insights into the training protocols necessary to develop robust pitch representations. However, the extension of reliable unsupervised competitive learning approaches from rate-coded to spiking neural networks has proven challenging, especially when biological plausibility and detail are desired. Transitioning to spiking neural network models is made more difficult by the comparatively high computational cost and complexity of these network models compared to rate-based models.</p> <p>We describe a transition from rate-based to spiking neural network models and address these outstanding issues. First, we focus on increasing simulation efficiency. We develop a state of the art graphical processing unit (GPU) based spiking neural network simulator which adopts optimisations from central processing unit (CPU) and cluster-based simulators. In benchmarks, we show that our novel simulator is capable of simulation speeds up to an order of magnitude greater than contemporary simulators. This greater simulation speed is intended to enable a higher throughput of spiking neural network simulations and thereby accelerate research. In order to ensure that present and future simulators can be efficiently and effectively compared, we also compile a repository of benchmarks which allow validation of simulator performance and simulated neural network dynamics.</p> <p>Having developed efficient simulation techniques, we propose a novel excitatory plasticity rule which, when used to train spiking neural networks in conjunction with an existing inhibitory plasticity rule, produces reliable competitive learning. These rules are both unsupervised, use local only information, and depend upon spike-timing in order to compute synaptic weight updates. This learning rule framework is used to produce a model of V1 simple cell receptive field development and shows qualitatively similar behaviour to traditional sparse coding models. During the development of these models we come across a number of challenges, especially in the interactions of inhibitory and excitatory neurons. We pose solutions to these challenges and thereafter suggest some future research questions and avenues for exploration.</p> |
| spellingShingle | Computational Neuroscience Ahmad, N Competitive learning in rate coded and spiking neural networks models with applications to vision and audition |
| title | Competitive learning in rate coded and spiking neural networks models with applications to vision and audition |
| title_full | Competitive learning in rate coded and spiking neural networks models with applications to vision and audition |
| title_fullStr | Competitive learning in rate coded and spiking neural networks models with applications to vision and audition |
| title_full_unstemmed | Competitive learning in rate coded and spiking neural networks models with applications to vision and audition |
| title_short | Competitive learning in rate coded and spiking neural networks models with applications to vision and audition |
| title_sort | competitive learning in rate coded and spiking neural networks models with applications to vision and audition |
| topic | Computational Neuroscience |
| work_keys_str_mv | AT ahmadn competitivelearninginratecodedandspikingneuralnetworksmodelswithapplicationstovisionandaudition |