{"id":{"repo_id":"denver","oai_identifier":"oai:digitalcommons.du.edu:etd-3084"},"canonical_url":"https://search.ndltd.org/etd/denver/oai:digitalcommons.du.edu:etd-3084","repository":{"repo_id":"denver","name":"University of Denver","base_url":"https://digitalcommons.du.edu/do/oai/"},"display":{"title":"Learning Approach for Fast Approximate Matrix Factorizations","abstract":"Efficiently computing an (approximate) orthonormal basis and low-rank approximation for the input data X plays a crucial role in data analysis. One of the most efficient algorithms for such tasks is the randomized algorithm, which proceeds by computing a projection XA with a random projection matrix A of much smaller size, and then computing the orthonormal basis as well as low-rank factorizations of the tall matrix XA. While a random matrix A is the de facto choice, in this work, we improve upon its performance by utilizing a learning approach to find an adaptive projection matrix A from a set of training data. We derive a closed-form formulation for the gradient of the training problem, enabling us to use efficient gradient-based algorithms. Experiments show that the learned dense matrix trained by eight different objective functions achieves better performance than a random one. We also extend this approach for learning structured projection matrix, such as the sketching matrix that performs as selecting a few number of representative columns from the input data. Our experiments on both synthetical and real data show that both learned dense and sketch projection matrices outperform the random ones in finding the approximate orthonormal basis and low-rank approximations. We conclude the thesis by discussing possible approaches for generalization analysis.","abstract_html":"Efficiently computing an (approximate) orthonormal basis and low-rank approximation for the input data X plays a crucial role in data analysis. One of the most efficient algorithms for such tasks is the randomized algorithm, which proceeds by computing a projection XA with a random projection matrix A of much smaller size, and then computing the orthonormal basis as well as low-rank factorizations of the tall matrix XA. While a random matrix A is the de facto choice, in this work, we improve upon its performance by utilizing a learning approach to find an adaptive projection matrix A from a set of training data. We derive a closed-form formulation for the gradient of the training problem, enabling us to use efficient gradient-based algorithms. Experiments show that the learned dense matrix trained by eight different objective functions achieves better performance than a random one. We also extend this approach for learning structured projection matrix, such as the sketching matrix that performs as selecting a few number of representative columns from the input data. Our experiments on both synthetical and real data show that both learned dense and sketch projection matrices outperform the random ones in finding the approximate orthonormal basis and low-rank approximations. We conclude the thesis by discussing possible approaches for generalization analysis.","abstract_has_math":false,"creators":["Yu, Haiyan"],"institution":null,"degree_name":"M.S.","degree_level":"Masters Thesis","degree_discipline":null,"degree_department":null,"school":null,"contributors":["Zhihui Zhu","Margareta Stefanovic","Rui Fan","Mei Yin"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":null,"date_issued":"","date_published":null,"updated_at":"2026-11-19T07:54:10Z","subjects":["Electrical engineering","Electrical and Computer Engineering","Engineering","Other Electrical and Computer Engineering"],"languages":["en"],"rights":["<p>Copyright is held by the author. User is responsible for all copyright compliance.</p>","Haiyan Yu"],"rights_urls":[],"identifier_entries":[],"person_identifier_entries":[]},"links":{"outbound_url":"https://digitalcommons.du.edu/etd/2091","outbound_label":"Repository record","outbound_source":"dc:identifier","record_url":"https://digitalcommons.du.edu/etd/2091","record_label":"Repository record","record_kind":"record","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://digitalcommons.du.edu/do/oai/?verb=GetRecord&metadataPrefix=oai_etdms&identifier=oai%3Adigitalcommons.du.edu%3Aetd-3084","prefix":"oai_etdms"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Zhihui Zhu","Margareta Stefanovic","Rui Fan","Mei Yin"]},{"key":"dc:creator","label":"Author","values":["Yu, Haiyan"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.available","label":"Dc Date Available","values":["2022-08-30T07:00:00Z"]},{"key":"dc:date.copyright","label":"Dc Date Copyright","values":["2022-01-01T08:00:00Z"]},{"key":"dc:date.created","label":"Dc Date Created","values":["2022-01-01T08:00:00Z"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Masters Thesis"]},{"key":"thesis:degree_name","label":"Degree Name","values":["M.S."]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Electrical engineering","Electrical and Computer Engineering","Engineering","Other Electrical and Computer Engineering"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["<p>Copyright is held by the author. User is responsible for all copyright compliance.</p>"]},{"key":"dc:rights.holder","label":"Dc Rights Holder","values":["Haiyan Yu"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://digitalcommons.du.edu/etd/2091"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Efficiently computing an (approximate) orthonormal basis and low-rank approximation for the input data X plays a crucial role in data analysis. One of the most efficient algorithms for such tasks is the randomized algorithm, which proceeds by computing a projection XA with a random projection matrix A of much smaller size, and then computing the orthonormal basis as well as low-rank factorizations of the tall matrix XA. While a random matrix A is the de facto choice, in this work, we improve upon its performance by utilizing a learning approach to find an adaptive projection matrix A from a set of training data. We derive a closed-form formulation for the gradient of the training problem, enabling us to use efficient gradient-based algorithms. Experiments show that the learned dense matrix trained by eight different objective functions achieves better performance than a random one. We also extend this approach for learning structured projection matrix, such as the sketching matrix that performs as selecting a few number of representative columns from the input data. Our experiments on both synthetical and real data show that both learned dense and sketch projection matrices outperform the random ones in finding the approximate orthonormal basis and low-rank approximations. We conclude the thesis by discussing possible approaches for generalization analysis."]},{"key":"dc:description.provenance","label":"Dc Description Provenance","values":["Received from ProQuest"]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:format.extent","label":"Dc Format Extent","values":["52 pgs"]},{"key":"dc:title","label":"Title","values":["Learning Approach for Fast Approximate Matrix Factorizations"]}]}],"canonical_facts":{"dc:contributor":["Zhihui Zhu","Margareta Stefanovic","Rui Fan","Mei Yin"],"dc:creator":["Yu, Haiyan"],"dc:date.available":["2022-08-30T07:00:00Z"],"dc:date.copyright":["2022-01-01T08:00:00Z"],"dc:date.created":["2022-01-01T08:00:00Z"],"dc:description.abstract":["Efficiently computing an (approximate) orthonormal basis and low-rank approximation for the input data X plays a crucial role in data analysis. One of the most efficient algorithms for such tasks is the randomized algorithm, which proceeds by computing a projection XA with a random projection matrix A of much smaller size, and then computing the orthonormal basis as well as low-rank factorizations of the tall matrix XA. While a random matrix A is the de facto choice, in this work, we improve upon its performance by utilizing a learning approach to find an adaptive projection matrix A from a set of training data. We derive a closed-form formulation for the gradient of the training problem, enabling us to use efficient gradient-based algorithms. Experiments show that the learned dense matrix trained by eight different objective functions achieves better performance than a random one. We also extend this approach for learning structured projection matrix, such as the sketching matrix that performs as selecting a few number of representative columns from the input data. Our experiments on both synthetical and real data show that both learned dense and sketch projection matrices outperform the random ones in finding the approximate orthonormal basis and low-rank approximations. We conclude the thesis by discussing possible approaches for generalization analysis."],"dc:description.provenance":["Received from ProQuest"],"dc:format":["application/pdf"],"dc:format.extent":["52 pgs"],"dc:identifier":["https://digitalcommons.du.edu/etd/2091"],"dc:language":["en"],"dc:rights":["<p>Copyright is held by the author. User is responsible for all copyright compliance.</p>"],"dc:rights.holder":["Haiyan Yu"],"dc:subject":["Electrical engineering","Electrical and Computer Engineering","Engineering","Other Electrical and Computer Engineering"],"dc:title":["Learning Approach for Fast Approximate Matrix Factorizations"],"thesis:degree_level":["Masters Thesis"],"thesis:degree_name":["M.S."]},"updated_at":"2026-11-19T07:54:10Z"}