{"id":{"repo_id":"rice","oai_identifier":"oai:repository.rice.edu:1911/105890"},"canonical_url":"https://search.dev.ndltd.org/etd/rice/oai:repository.rice.edu:1911/105890","repository":{"repo_id":"rice","name":"Rice University","base_url":"https://repository.rice.edu/server/oai/request"},"display":{"title":"Phase Retrieval Under a Generative Prior","abstract":"The phase retrieval problem, arising from X-ray crystallography and medical imaging, asks to recover a signal given intensity-only measurements. When the number of measurements is less than the dimensionality of the signal, solving the problem requires additional assumptions, or priors, on its structure in order to guarantee recovery. Many techniques enforce a sparsity prior, meaning that the signal has very few non-zero entries. However, these methods have seen various computational bottlenecks. We sidestep this issue by enforcing a generative prior: the assumption that the signal is in the range of a generative neural network. By formulating an empirical risk minimization problem and directly optimizing over the domain of the generator, we show that the objective’s energy landscape exhibits favorable global geometry for gradient descent with information theoretically optimal sample complexity. Based on this geometric result, we introduce a gradient descent algorithm to converge to the true solution. We corroborate these results with experiments showing that exploiting generative models in phase retrieval tasks outperforms sparse phase retrieval methods.","abstract_html":"The phase retrieval problem, arising from X-ray crystallography and medical imaging, asks to recover a signal given intensity-only measurements. When the number of measurements is less than the dimensionality of the signal, solving the problem requires additional assumptions, or priors, on its structure in order to guarantee recovery. Many techniques enforce a sparsity prior, meaning that the signal has very few non-zero entries. However, these methods have seen various computational bottlenecks. We sidestep this issue by enforcing a generative prior: the assumption that the signal is in the range of a generative neural network. By formulating an empirical risk minimization problem and directly optimizing over the domain of the generator, we show that the objective’s energy landscape exhibits favorable global geometry for gradient descent with information theoretically optimal sample complexity. Based on this geometric result, we introduce a gradient descent algorithm to converge to the true solution. We corroborate these results with experiments showing that exploiting generative models in phase retrieval tasks outperforms sparse phase retrieval methods.","abstract_has_math":false,"creators":["Leong, Oscar"],"institution":"Rice University","degree_name":"Master of Arts","degree_level":"Masters","degree_discipline":"Engineering","degree_department":null,"school":null,"contributors":[],"advisors":["Hicks, Illya"],"committee_chairs":[],"committee_members":["Hand, Paul"],"year":2019,"date_issued":"2019-04-11","date_published":"2019-04-11","updated_at":"2026-07-24T04:10:24Z","subjects":["Phase Retrieval","Generative Models","Non-convex Optimization","Deep Learning"],"languages":["eng"],"rights":["Copyright is held by the author, unless otherwise indicated. Permission to reuse, publish, or reproduce the work beyond the bounds of fair use or other exemptions to copyright law must be obtained from the copyright holder."],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/1911/105890","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Hicks, Illya"]},{"key":"dc:contributor.committeemember","label":"Committee Member","values":["Hand, Paul"]},{"key":"dc:creator","label":"Author","values":["Leong, Oscar"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2019-05-17T16:45:56Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2019-05-17T16:45:56Z"]},{"key":"dc:date.issued","label":"Date","values":["2019-04-11"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Engineering"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Masters"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master of Arts"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Rice University"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Phase Retrieval","Generative Models","Non-convex Optimization","Deep Learning"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["Copyright is held by the author, unless otherwise indicated. Permission to reuse, publish, or reproduce the work beyond the bounds of fair use or other exemptions to copyright law must be obtained from the copyright holder."]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/1911/105890"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["The phase retrieval problem, arising from X-ray crystallography and medical imaging, asks to recover a signal given intensity-only measurements. When the number of measurements is less than the dimensionality of the signal, solving the problem requires additional assumptions, or priors, on its structure in order to guarantee recovery. Many techniques enforce a sparsity prior, meaning that the signal has very few non-zero entries. However, these methods have seen various computational bottlenecks. We sidestep this issue by enforcing a generative prior: the assumption that the signal is in the range of a generative neural network. By formulating an empirical risk minimization problem and directly optimizing over the domain of the generator, we show that the objective’s energy landscape exhibits favorable global geometry for gradient descent with information theoretically optimal sample complexity. Based on this geometric result, we introduce a gradient descent algorithm to converge to the true solution. We corroborate these results with experiments showing that exploiting generative models in phase retrieval tasks outperforms sparse phase retrieval methods."]},{"key":"dc:format.mimetype","label":"Dc Format Mimetype","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Phase Retrieval Under a Generative Prior"]}]}],"canonical_facts":{"dc:contributor.advisor":["Hicks, Illya"],"dc:contributor.committeemember":["Hand, Paul"],"dc:creator":["Leong, Oscar"],"dc:date.accessioned":["2019-05-17T16:45:56Z"],"dc:date.available":["2019-05-17T16:45:56Z"],"dc:date.issued":["2019-04-11"],"dc:description.abstract":["The phase retrieval problem, arising from X-ray crystallography and medical imaging, asks to recover a signal given intensity-only measurements. When the number of measurements is less than the dimensionality of the signal, solving the problem requires additional assumptions, or priors, on its structure in order to guarantee recovery. Many techniques enforce a sparsity prior, meaning that the signal has very few non-zero entries. However, these methods have seen various computational bottlenecks. We sidestep this issue by enforcing a generative prior: the assumption that the signal is in the range of a generative neural network. By formulating an empirical risk minimization problem and directly optimizing over the domain of the generator, we show that the objective’s energy landscape exhibits favorable global geometry for gradient descent with information theoretically optimal sample complexity. Based on this geometric result, we introduce a gradient descent algorithm to converge to the true solution. We corroborate these results with experiments showing that exploiting generative models in phase retrieval tasks outperforms sparse phase retrieval methods."],"dc:format.mimetype":["application/pdf"],"dc:identifier.uri":["https://hdl.handle.net/1911/105890"],"dc:language.iso":["eng"],"dc:rights":["Copyright is held by the author, unless otherwise indicated. Permission to reuse, publish, or reproduce the work beyond the bounds of fair use or other exemptions to copyright law must be obtained from the copyright holder."],"dc:subject":["Phase Retrieval","Generative Models","Non-convex Optimization","Deep Learning"],"dc:title":["Phase Retrieval Under a Generative Prior"],"dc:type":["Thesis"],"thesis:degree_discipline":["Engineering"],"thesis:degree_level":["Masters"],"thesis:degree_name":["Master of Arts"],"thesis:institution_name":["Rice University"]},"updated_at":"2026-07-24T04:10:24Z"}