{"id":{"repo_id":"cambridge","oai_identifier":"oai:www.repository.cam.ac.uk:1810/386727"},"canonical_url":"https://search.dev.ndltd.org/etd/cambridge/oai:www.repository.cam.ac.uk:1810/386727","repository":{"repo_id":"cambridge","name":"Cambridge University","base_url":"https://api.repository.cam.ac.uk/server/oai/request"},"display":{"title":"Oscillatory Spiking Circuits: A biologically plausible architecture for fast and reliable spiking computations","abstract":"The functions carried out by the brain are completed in a large part by dynamically complicated neurons which transmit information in discrete pulses, called spikes. The brain uses these to compute in a (usually) reliable fashion, and can do so rapidly. How to construct Spiking Neural Networks (SNNs) that achieve similar performance is still unknown. Here, we present an architecture for spiking neural networks that operate efficiently (with a low number of spikes), rapidly, and robustly, using biologically plausible components and operating within biological constraints. At the heart lie neuronal oscillations which are produced by recurrent excitation and inhibition, called PING. The oscillations organise the spikes in time, putting the circuit in a binary regime, with each neuron having an integrating time window that can accommodate at most one spike from each upstream neuron. Firstly, we investigate how naive approaches to construct SNNs can fail when confronted with spike timing jitter. We then propose a computation role for dendritic action potentials with long time constants. Such dendritic spikes exhibit long plateau potentials which can last tens of milliseconds, outliving somatic spikes by an order of magnitude. We propose that this long time constant allow for the reliable integration of asynchronous inputs. Then we turn our attention to PING rhythms. We briefly look how they arise and show that these rhythms remain reliable in the face of disturbances and parameter uncertainty, and introduce a tool to investigate feedback on the level of the network: the population voltage clamp. Using this, we demonstrate that PING is reliable due to slow positive feedback arising from synaptic interactions. Finally, we make the case that PING circuits can indeed support a binary computation. After examining why gradient-based methods can struggle in the training of spiking systems, we introduce a new method of training biological neural networks, which we call teacher- forcing.","abstract_html":"The functions carried out by the brain are completed in a large part by dynamically complicated neurons which transmit information in discrete pulses, called spikes. The brain uses these to compute in a (usually) reliable fashion, and can do so rapidly. How to construct Spiking Neural Networks (SNNs) that achieve similar performance is still unknown. Here, we present an architecture for spiking neural networks that operate efficiently (with a low number of spikes), rapidly, and robustly, using biologically plausible components and operating within biological constraints. At the heart lie neuronal oscillations which are produced by recurrent excitation and inhibition, called PING. The oscillations organise the spikes in time, putting the circuit in a binary regime, with each neuron having an integrating time window that can accommodate at most one spike from each upstream neuron. Firstly, we investigate how naive approaches to construct SNNs can fail when confronted with spike timing jitter. We then propose a computation role for dendritic action potentials with long time constants. Such dendritic spikes exhibit long plateau potentials which can last tens of milliseconds, outliving somatic spikes by an order of magnitude. We propose that this long time constant allow for the reliable integration of asynchronous inputs. Then we turn our attention to PING rhythms. We briefly look how they arise and show that these rhythms remain reliable in the face of disturbances and parameter uncertainty, and introduce a tool to investigate feedback on the level of the network: the population voltage clamp. Using this, we demonstrate that PING is reliable due to slow positive feedback arising from synaptic interactions. Finally, we make the case that PING circuits can indeed support a binary computation. After examining why gradient-based methods can struggle in the training of spiking systems, we introduce a new method of training biological neural networks, which we call teacher- forcing.","abstract_has_math":false,"creators":["Burger, Thomas"],"institution":"University of Cambridge","degree_name":"Doctor of Philosophy (PhD)","degree_level":"Doctoral","degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":["O'Leary, Timothy"],"committee_chairs":[],"committee_members":[],"year":2024,"date_issued":"2024-01-26","date_published":"2024-01-26","updated_at":"2026-07-22T22:24:28Z","subjects":["neuromorphic computing","neurophysiology","spiking neural networks"],"languages":["eng"],"rights":[],"rights_urls":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/be663389-11ab-4519-bdcc-dd2f80c3773a/download","https://creativecommons.org/licenses/by/4.0/"],"identifier_entries":[]},"links":{"outbound_url":"https://doi.org/10.17863/CAM.119821","outbound_label":"DOI","outbound_source":"dc:identifier.doi"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["O'Leary, Timothy"]},{"key":"dc:creator","label":"Author","values":["Burger, Thomas"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.issued","label":"Date","values":["2024-01-26"]},{"key":"dc:publisher.institution","label":"Dc Publisher Institution","values":["University of Cambridge"]},{"key":"dc:relation.isreferencedby.uri","label":"Dc Relation Isreferencedby URI","values":["https://www.repository.cam.ac.uk/handle/1810/386727"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"dc:type.qualificationlevel","label":"Dc Type Qualificationlevel","values":["Doctoral"]},{"key":"dc:type.qualificationname","label":"Dc Type Qualificationname","values":["Doctor of Philosophy (PhD)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["neuromorphic computing","neurophysiology","spiking neural networks"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/be663389-11ab-4519-bdcc-dd2f80c3773a/download","https://creativecommons.org/licenses/by/4.0/"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.doi","label":"DOI","values":["https://doi.org/10.17863/CAM.119821"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/ef26622e-e7d4-42f3-9f0b-8274756afea4/download"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["The functions carried out by the brain are completed in a large part by dynamically complicated neurons which transmit information in discrete pulses, called spikes. The brain uses these to compute in a (usually) reliable fashion, and can do so rapidly. How to construct Spiking Neural Networks (SNNs) that achieve similar performance is still unknown. Here, we present an architecture for spiking neural networks that operate efficiently (with a low number of spikes), rapidly, and robustly, using biologically plausible components and operating within biological constraints. At the heart lie neuronal oscillations which are produced by recurrent excitation and inhibition, called PING. The oscillations organise the spikes in time, putting the circuit in a binary regime, with each neuron having an integrating time window that can accommodate at most one spike from each upstream neuron. Firstly, we investigate how naive approaches to construct SNNs can fail when confronted with spike timing jitter. We then propose a computation role for dendritic action potentials with long time constants. Such dendritic spikes exhibit long plateau potentials which can last tens of milliseconds, outliving somatic spikes by an order of magnitude. We propose that this long time constant allow for the reliable integration of asynchronous inputs. Then we turn our attention to PING rhythms. We briefly look how they arise and show that these rhythms remain reliable in the face of disturbances and parameter uncertainty, and introduce a tool to investigate feedback on the level of the network: the population voltage clamp. Using this, we demonstrate that PING is reliable due to slow positive feedback arising from synaptic interactions. Finally, we make the case that PING circuits can indeed support a binary computation. After examining why gradient-based methods can struggle in the training of spiking systems, we introduce a new method of training biological neural networks, which we call teacher- forcing."]},{"key":"dc:format.checksum.md5","label":"Dc Format Checksum Md5","values":["7be3f30b78aaee91531bd9f32eb73403","87eda9de84448d1f82354d60eee3eb5f"]},{"key":"dc:title","label":"Title","values":["Oscillatory Spiking Circuits: A biologically plausible architecture for fast and reliable spiking computations"]}]}],"canonical_facts":{"dc:contributor.advisor":["O'Leary, Timothy"],"dc:creator":["Burger, Thomas"],"dc:date.issued":["2024-01-26"],"dc:description.abstract":["The functions carried out by the brain are completed in a large part by dynamically complicated neurons which transmit information in discrete pulses, called spikes. The brain uses these to compute in a (usually) reliable fashion, and can do so rapidly. How to construct Spiking Neural Networks (SNNs) that achieve similar performance is still unknown. Here, we present an architecture for spiking neural networks that operate efficiently (with a low number of spikes), rapidly, and robustly, using biologically plausible components and operating within biological constraints. At the heart lie neuronal oscillations which are produced by recurrent excitation and inhibition, called PING. The oscillations organise the spikes in time, putting the circuit in a binary regime, with each neuron having an integrating time window that can accommodate at most one spike from each upstream neuron. Firstly, we investigate how naive approaches to construct SNNs can fail when confronted with spike timing jitter. We then propose a computation role for dendritic action potentials with long time constants. Such dendritic spikes exhibit long plateau potentials which can last tens of milliseconds, outliving somatic spikes by an order of magnitude. We propose that this long time constant allow for the reliable integration of asynchronous inputs. Then we turn our attention to PING rhythms. We briefly look how they arise and show that these rhythms remain reliable in the face of disturbances and parameter uncertainty, and introduce a tool to investigate feedback on the level of the network: the population voltage clamp. Using this, we demonstrate that PING is reliable due to slow positive feedback arising from synaptic interactions. Finally, we make the case that PING circuits can indeed support a binary computation. After examining why gradient-based methods can struggle in the training of spiking systems, we introduce a new method of training biological neural networks, which we call teacher- forcing."],"dc:format.checksum.md5":["7be3f30b78aaee91531bd9f32eb73403","87eda9de84448d1f82354d60eee3eb5f"],"dc:identifier.doi":["https://doi.org/10.17863/CAM.119821"],"dc:identifier.uri":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/ef26622e-e7d4-42f3-9f0b-8274756afea4/download"],"dc:language":["eng"],"dc:publisher.institution":["University of Cambridge"],"dc:relation.isreferencedby.uri":["https://www.repository.cam.ac.uk/handle/1810/386727"],"dc:rights":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/be663389-11ab-4519-bdcc-dd2f80c3773a/download","https://creativecommons.org/licenses/by/4.0/"],"dc:subject":["neuromorphic computing","neurophysiology","spiking neural networks"],"dc:title":["Oscillatory Spiking Circuits: A biologically plausible architecture for fast and reliable spiking computations"],"dc:type":["Thesis"],"dc:type.qualificationlevel":["Doctoral"],"dc:type.qualificationname":["Doctor of Philosophy (PhD)"]},"updated_at":"2026-07-22T22:24:28Z"}