{"id":{"repo_id":"cambridge","oai_identifier":"oai:www.repository.cam.ac.uk:1810/379820"},"canonical_url":"https://search.dev.ndltd.org/etd/cambridge/oai:www.repository.cam.ac.uk:1810/379820","repository":{"repo_id":"cambridge","name":"Cambridge University","base_url":"https://api.repository.cam.ac.uk/server/oai/request"},"display":{"title":"Estimating the Performance of Optical Fibre Communication Systems","abstract":"The growth in demand for reliable and economically viable optical communication systems has increased the research into improving their capacity. In light of the wide availability of data generated by these systems and the complexity of their operation given the associated non-linearities, we employ machine learning methods to complement analytical solutions of the physical phenomena and mathematical concepts involved. The work in this thesis is focused at the receiver end, following the digital signal processing phase. First, we reconsider the application of the multicanonical Monte Carlo method to estimate the very low bit error rate of low-density parity-check codes, all while employing a parallel belief-propagation decoder implementation. Additionally, we look into the accuracy and practicality of such implementations. Second, we consider performance improvement of these error correction codes using weighted belief propagation that targets problematic sets found in their Tanner graph representation, and called trapping sets. The framework involves first locating the trapping sets, then determining the ones that result in the highest error rates, and lastly marking the relevant Tanner graph edges for belief propagation weight change. The corresponding weights are determined using machine learning given the high structural complexity of these codes. Lastly, and given that the performance of any forward error correction code is contingent on the accurate estimation of the signal-to-noise ratio, we look into accurately estimating the latter using hybrid models which not only require less data than machine learning models, but are also interpretable. The hybrid models consist of a measurement-informed physical model, developed by systematically reducing the number of independent parameters based on the underpinning physics, namely the Gaussian noise model. It is then integrated with machine learning, specifically Gaussian process regression given its ability for accurate uncertainty estimation, in two different hybrid models to further decrease the error margin in evaluating the signal-to-noise ratio and account for phenomena not expressed in the physical model. We compare the accuracy of the estimations using these models to ones from data-driven approaches such as neural networks and Gaussian process regression. Planned future works will focus on suggestions for extending the presented frameworks to incorporate machine learning techniques and hardware technological advancements.","abstract_html":"The growth in demand for reliable and economically viable optical communication systems has increased the research into improving their capacity. In light of the wide availability of data generated by these systems and the complexity of their operation given the associated non-linearities, we employ machine learning methods to complement analytical solutions of the physical phenomena and mathematical concepts involved. The work in this thesis is focused at the receiver end, following the digital signal processing phase. First, we reconsider the application of the multicanonical Monte Carlo method to estimate the very low bit error rate of low-density parity-check codes, all while employing a parallel belief-propagation decoder implementation. Additionally, we look into the accuracy and practicality of such implementations. Second, we consider performance improvement of these error correction codes using weighted belief propagation that targets problematic sets found in their Tanner graph representation, and called trapping sets. The framework involves first locating the trapping sets, then determining the ones that result in the highest error rates, and lastly marking the relevant Tanner graph edges for belief propagation weight change. The corresponding weights are determined using machine learning given the high structural complexity of these codes. Lastly, and given that the performance of any forward error correction code is contingent on the accurate estimation of the signal-to-noise ratio, we look into accurately estimating the latter using hybrid models which not only require less data than machine learning models, but are also interpretable. The hybrid models consist of a measurement-informed physical model, developed by systematically reducing the number of independent parameters based on the underpinning physics, namely the Gaussian noise model. It is then integrated with machine learning, specifically Gaussian process regression given its ability for accurate uncertainty estimation, in two different hybrid models to further decrease the error margin in evaluating the signal-to-noise ratio and account for phenomena not expressed in the physical model. We compare the accuracy of the estimations using these models to ones from data-driven approaches such as neural networks and Gaussian process regression. Planned future works will focus on suggestions for extending the presented frameworks to incorporate machine learning techniques and hardware technological advancements.","abstract_has_math":false,"creators":["Mansour, Mariane"],"institution":"University of Cambridge","degree_name":"Doctor of Philosophy (PhD)","degree_level":"Doctoral","degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":["Savory, Seb"],"committee_chairs":[],"committee_members":[],"year":2024,"date_issued":"2024-10-06","date_published":"2024-10-06","updated_at":"2026-07-22T22:24:17Z","subjects":["Gaussian Process","LDPC","Machine Learning","Optical Communication"],"languages":["eng"],"rights":[],"rights_urls":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/1786f159-9433-4e2e-ac3c-fadebd59a3da/download","http://purl.org/NET/rdflicense/allrightsreserved"],"identifier_entries":[]},"links":{"outbound_url":"https://doi.org/10.17863/CAM.115791","outbound_label":"DOI","outbound_source":"dc:identifier.doi"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Savory, Seb"]},{"key":"dc:contributor.sponsor","label":"Sponsor","values":["This thesis was funded by Ciena."]},{"key":"dc:creator","label":"Author","values":["Mansour, Mariane"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.issued","label":"Date","values":["2024-10-06"]},{"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/1810/379820"]},{"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":["Gaussian Process","LDPC","Machine Learning","Optical Communication"]}]},{"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/1786f159-9433-4e2e-ac3c-fadebd59a3da/download","http://purl.org/NET/rdflicense/allrightsreserved"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.doi","label":"DOI","values":["https://doi.org/10.17863/CAM.115791"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/34cb7ab0-b36b-4e4f-9096-b09d6fd7534b/download"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["The growth in demand for reliable and economically viable optical communication systems has increased the research into improving their capacity. In light of the wide availability of data generated by these systems and the complexity of their operation given the associated non-linearities, we employ machine learning methods to complement analytical solutions of the physical phenomena and mathematical concepts involved. The work in this thesis is focused at the receiver end, following the digital signal processing phase. First, we reconsider the application of the multicanonical Monte Carlo method to estimate the very low bit error rate of low-density parity-check codes, all while employing a parallel belief-propagation decoder implementation. Additionally, we look into the accuracy and practicality of such implementations. Second, we consider performance improvement of these error correction codes using weighted belief propagation that targets problematic sets found in their Tanner graph representation, and called trapping sets. The framework involves first locating the trapping sets, then determining the ones that result in the highest error rates, and lastly marking the relevant Tanner graph edges for belief propagation weight change. The corresponding weights are determined using machine learning given the high structural complexity of these codes. Lastly, and given that the performance of any forward error correction code is contingent on the accurate estimation of the signal-to-noise ratio, we look into accurately estimating the latter using hybrid models which not only require less data than machine learning models, but are also interpretable. The hybrid models consist of a measurement-informed physical model, developed by systematically reducing the number of independent parameters based on the underpinning physics, namely the Gaussian noise model. It is then integrated with machine learning, specifically Gaussian process regression given its ability for accurate uncertainty estimation, in two different hybrid models to further decrease the error margin in evaluating the signal-to-noise ratio and account for phenomena not expressed in the physical model. We compare the accuracy of the estimations using these models to ones from data-driven approaches such as neural networks and Gaussian process regression. Planned future works will focus on suggestions for extending the presented frameworks to incorporate machine learning techniques and hardware technological advancements."]},{"key":"dc:format.checksum.md5","label":"Dc Format Checksum Md5","values":["e97057d04bd33c73ff914d8fae270af2","87eda9de84448d1f82354d60eee3eb5f"]},{"key":"dc:title","label":"Title","values":["Estimating the Performance of Optical Fibre Communication Systems"]}]}],"canonical_facts":{"dc:contributor.advisor":["Savory, Seb"],"dc:contributor.sponsor":["This thesis was funded by Ciena."],"dc:creator":["Mansour, Mariane"],"dc:date.issued":["2024-10-06"],"dc:description.abstract":["The growth in demand for reliable and economically viable optical communication systems has increased the research into improving their capacity. In light of the wide availability of data generated by these systems and the complexity of their operation given the associated non-linearities, we employ machine learning methods to complement analytical solutions of the physical phenomena and mathematical concepts involved. The work in this thesis is focused at the receiver end, following the digital signal processing phase. First, we reconsider the application of the multicanonical Monte Carlo method to estimate the very low bit error rate of low-density parity-check codes, all while employing a parallel belief-propagation decoder implementation. Additionally, we look into the accuracy and practicality of such implementations. Second, we consider performance improvement of these error correction codes using weighted belief propagation that targets problematic sets found in their Tanner graph representation, and called trapping sets. The framework involves first locating the trapping sets, then determining the ones that result in the highest error rates, and lastly marking the relevant Tanner graph edges for belief propagation weight change. The corresponding weights are determined using machine learning given the high structural complexity of these codes. Lastly, and given that the performance of any forward error correction code is contingent on the accurate estimation of the signal-to-noise ratio, we look into accurately estimating the latter using hybrid models which not only require less data than machine learning models, but are also interpretable. The hybrid models consist of a measurement-informed physical model, developed by systematically reducing the number of independent parameters based on the underpinning physics, namely the Gaussian noise model. It is then integrated with machine learning, specifically Gaussian process regression given its ability for accurate uncertainty estimation, in two different hybrid models to further decrease the error margin in evaluating the signal-to-noise ratio and account for phenomena not expressed in the physical model. We compare the accuracy of the estimations using these models to ones from data-driven approaches such as neural networks and Gaussian process regression. Planned future works will focus on suggestions for extending the presented frameworks to incorporate machine learning techniques and hardware technological advancements."],"dc:format.checksum.md5":["e97057d04bd33c73ff914d8fae270af2","87eda9de84448d1f82354d60eee3eb5f"],"dc:identifier.doi":["https://doi.org/10.17863/CAM.115791"],"dc:identifier.uri":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/34cb7ab0-b36b-4e4f-9096-b09d6fd7534b/download"],"dc:language":["eng"],"dc:publisher.institution":["University of Cambridge"],"dc:relation.isreferencedby.uri":["https://www.repository.cam.ac.uk/handle/1810/379820"],"dc:rights":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/1786f159-9433-4e2e-ac3c-fadebd59a3da/download","http://purl.org/NET/rdflicense/allrightsreserved"],"dc:subject":["Gaussian Process","LDPC","Machine Learning","Optical Communication"],"dc:title":["Estimating the Performance of Optical Fibre Communication Systems"],"dc:type":["Thesis"],"dc:type.qualificationlevel":["Doctoral"],"dc:type.qualificationname":["Doctor of Philosophy (PhD)"]},"updated_at":"2026-07-22T22:24:17Z"}