{"id":{"repo_id":"buffalo","oai_identifier":"oai:ubir.buffalo.edu:10477/79361"},"canonical_url":"https://search.dev.ndltd.org/etd/buffalo/oai:ubir.buffalo.edu:10477/79361","repository":{"repo_id":"buffalo","name":"Buffalo","base_url":"https://ubir.buffalo.edu/oai/request"},"display":{"title":"Application of the Tensor Train Decomposition in Machine Learning - A Study and Tradeoffs","abstract":"M.S.","abstract_html":"M.S.","abstract_has_math":false,"creators":["Wilkinson, Andrew; 0000-0002-0237-1775"],"institution":"State University of New York at Buffalo","degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Sridhar, Ramalingam","Computer Science and Engineering"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2019,"date_issued":"2019-04-04T20:30:50Z","date_published":"2019-04-04T20:30:50Z","updated_at":"2026-07-27T19:05:16Z","subjects":["computer engineering"],"languages":["eng"],"rights":["Users of works found in University at Buffalo Institutional Repository (UBIR) are responsible for identifying and contacting the copyright owner for permission to reuse. University at Buffalo Libraries do not manage rights for copyright-protected works and cannot assist with permissions.","Copyright retained by author."],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/10477/79361","outbound_label":"Handle","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Sridhar, Ramalingam","Computer Science and Engineering"]},{"key":"dc:creator","label":"Author","values":["Wilkinson, Andrew; 0000-0002-0237-1775"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2019-04-04T20:30:50Z","2019","2018-12-26 13:50:12"]},{"key":"dc:publisher","label":"Institution","values":["State University of New York at Buffalo"]},{"key":"dc:type","label":"Dc Type","values":["Text","Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["computer engineering"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["Users of works found in University at Buffalo Institutional Repository (UBIR) are responsible for identifying and contacting the copyright owner for permission to reuse. University at Buffalo Libraries do not manage rights for copyright-protected works and cannot assist with permissions.","Copyright retained by author."]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["http://hdl.handle.net/10477/79361"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["M.S.","Modern applications in machine learning are increasingly focused on large and distributed sets of data, which are often multidimensional in nature. Such multidimensional datasets are referred to as tensors. Developing efficient methods and algorithms for working with, and learning from these datasets has been the focus of recent of research. Tensor decompositions provide efficient methods for representing large tensors in more compact formats, and has opened the door for a wide variety of applications. In this thesis, we examine several methods by which such tensor decompositions can be used to improve machine learning, from both software and hardware perspectives. Specifically, we examine the Tensor Train (TT) and Quantized-Tensor Train (QTT) decompositions, and their applications for machine learning. Machine learning applications involving the TT decomposition have been used widely for the purposes of compressing Neural Network (NN) weights. The TT decomposition can significantly compress these networks, which in some instances contain over 100 million nodes, while still maintaining the expressive power of their layers. Here, we take the novel approach of applying the TT decomposition to NN inputs, and using these compressed inputs to train and test the NN model. We then compare this approach to other methods of feature reduction such as Principal Component Analysis (PCA) and streaming PCA."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Application of the Tensor Train Decomposition in Machine Learning - A Study and Tradeoffs"]}]}],"canonical_facts":{"dc:contributor":["Sridhar, Ramalingam","Computer Science and Engineering"],"dc:creator":["Wilkinson, Andrew; 0000-0002-0237-1775"],"dc:date":["2019-04-04T20:30:50Z","2019","2018-12-26 13:50:12"],"dc:description":["M.S.","Modern applications in machine learning are increasingly focused on large and distributed sets of data, which are often multidimensional in nature. Such multidimensional datasets are referred to as tensors. Developing efficient methods and algorithms for working with, and learning from these datasets has been the focus of recent of research. Tensor decompositions provide efficient methods for representing large tensors in more compact formats, and has opened the door for a wide variety of applications. In this thesis, we examine several methods by which such tensor decompositions can be used to improve machine learning, from both software and hardware perspectives. Specifically, we examine the Tensor Train (TT) and Quantized-Tensor Train (QTT) decompositions, and their applications for machine learning. Machine learning applications involving the TT decomposition have been used widely for the purposes of compressing Neural Network (NN) weights. The TT decomposition can significantly compress these networks, which in some instances contain over 100 million nodes, while still maintaining the expressive power of their layers. Here, we take the novel approach of applying the TT decomposition to NN inputs, and using these compressed inputs to train and test the NN model. We then compare this approach to other methods of feature reduction such as Principal Component Analysis (PCA) and streaming PCA."],"dc:format":["application/pdf"],"dc:identifier":["http://hdl.handle.net/10477/79361"],"dc:language":["eng"],"dc:publisher":["State University of New York at Buffalo"],"dc:rights":["Users of works found in University at Buffalo Institutional Repository (UBIR) are responsible for identifying and contacting the copyright owner for permission to reuse. University at Buffalo Libraries do not manage rights for copyright-protected works and cannot assist with permissions.","Copyright retained by author."],"dc:subject":["computer engineering"],"dc:title":["Application of the Tensor Train Decomposition in Machine Learning - A Study and Tradeoffs"],"dc:type":["Text","Thesis"]},"updated_at":"2026-07-27T19:05:16Z"}