{"id":{"repo_id":"uiuc","oai_identifier":"oai:www.ideals.illinois.edu:2142/100951"},"canonical_url":"https://search.dev.ndltd.org/etd/uiuc/oai:www.ideals.illinois.edu:2142/100951","repository":{"repo_id":"uiuc","name":"University of Illinois - Urbana-Champaign","base_url":"https://www.ideals.illinois.edu/oai-pmh"},"display":{"title":"Generative modeling using the sliced Wasserstein distance","abstract":"Generative adversarial nets (GANs) are very successful at modeling distributions from given samples, even in the high-dimensional case. However, their formulation is also known to be hard to optimize and often unstable. While the aforementioned problems are particularly true for early GAN formulations, there has been significant empirically motivated and theoretically founded progress to improve stability, for instance, by using the Wasserstein distance rather than the Jenson-Shannon divergence. Here, we consider an alternative formulation for generative modeling based on random projections which, in its simplest form, results in a single objective rather than a saddlepoint formulation. By augmenting this approach with a discriminator we improve its accuracy. We found our approach to be significantly more stable compared to even the improved Wasserstein GAN. Further, unlike the traditional GAN loss, the loss formulated in our method is a good measure of the actual distance between the distributions and, for the first time for GAN training, we are able to show estimates for the same.","abstract_html":"Generative adversarial nets (GANs) are very successful at modeling distributions from given samples, even in the high-dimensional case. However, their formulation is also known to be hard to optimize and often unstable. While the aforementioned problems are particularly true for early GAN formulations, there has been significant empirically motivated and theoretically founded progress to improve stability, for instance, by using the Wasserstein distance rather than the Jenson-Shannon divergence. Here, we consider an alternative formulation for generative modeling based on random projections which, in its simplest form, results in a single objective rather than a saddlepoint formulation. By augmenting this approach with a discriminator we improve its accuracy. We found our approach to be significantly more stable compared to even the improved Wasserstein GAN. Further, unlike the traditional GAN loss, the loss formulated in our method is a good measure of the actual distance between the distributions and, for the first time for GAN training, we are able to show estimates for the same.","abstract_has_math":false,"creators":["Deshpande, Ishan"],"institution":"University of Illinois at Urbana-Champaign","degree_name":"M.S.","degree_level":"Thesis","degree_discipline":"Electrical & Computer Engr","degree_department":null,"school":null,"contributors":["Schwing, Alexander G."],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2018,"date_issued":"2018-09-04T20:26:58Z","date_published":"2018-09-04T20:26:58Z","updated_at":"2026-07-22T22:24:38Z","subjects":["Machine Learning","Deep Learning"],"languages":["en"],"rights":["Copyright 2018 Ishan Deshpande"],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/2142/100951","outbound_label":"Handle","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Schwing, Alexander G."]},{"key":"dc:creator","label":"Author","values":["Deshpande, Ishan"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2018-09-04T20:26:58Z","2018-04-11","2018-05"]},{"key":"dc:type","label":"Dc Type","values":["text"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Electrical & Computer Engr"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Thesis"]},{"key":"thesis:degree_name","label":"Degree Name","values":["M.S."]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["University of Illinois at Urbana-Champaign"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Machine Learning","Deep Learning"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["Copyright 2018 Ishan Deshpande"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["http://hdl.handle.net/2142/100951"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Generative adversarial nets (GANs) are very successful at modeling distributions from given samples, even in the high-dimensional case. However, their formulation is also known to be hard to optimize and often unstable. While the aforementioned problems are particularly true for early GAN formulations, there has been significant empirically motivated and theoretically founded progress to improve stability, for instance, by using the Wasserstein distance rather than the Jenson-Shannon divergence. Here, we consider an alternative formulation for generative modeling based on random projections which, in its simplest form, results in a single objective rather than a saddlepoint formulation. By augmenting this approach with a discriminator we improve its accuracy. We found our approach to be significantly more stable compared to even the improved Wasserstein GAN. Further, unlike the traditional GAN loss, the loss formulated in our method is a good measure of the actual distance between the distributions and, for the first time for GAN training, we are able to show estimates for the same.","Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2018-08-31 without embargo terms","The student, Ishan Deshpande, accepted the attached license on 2018-04-10 at 13:35.","The student, Ishan Deshpande, submitted this Thesis for approval on 2018-04-10 at 13:40.","This Thesis was approved for publication on 2018-04-11 at 14:56.","DSpace SAF Submission Ingestion Package generated from Vireo submission #12168 on 2018-08-31 at 17:11:38","Made available in DSpace on 2018-09-04T20:26:58Z (GMT). No. of bitstreams: 2 DESHPANDE-THESIS-2018.pdf: 14373491 bytes, checksum: beb5abc56a3c12e9d138b2f9708b96bb (MD5) LICENSE.txt: 4212 bytes, checksum: 210f1676a1810002500290549bf8041a (MD5) Previous issue date: 2018-04-11"]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Generative modeling using the sliced Wasserstein distance"]}]}],"canonical_facts":{"dc:contributor":["Schwing, Alexander G."],"dc:creator":["Deshpande, Ishan"],"dc:date":["2018-09-04T20:26:58Z","2018-04-11","2018-05"],"dc:description":["Generative adversarial nets (GANs) are very successful at modeling distributions from given samples, even in the high-dimensional case. However, their formulation is also known to be hard to optimize and often unstable. While the aforementioned problems are particularly true for early GAN formulations, there has been significant empirically motivated and theoretically founded progress to improve stability, for instance, by using the Wasserstein distance rather than the Jenson-Shannon divergence. Here, we consider an alternative formulation for generative modeling based on random projections which, in its simplest form, results in a single objective rather than a saddlepoint formulation. By augmenting this approach with a discriminator we improve its accuracy. We found our approach to be significantly more stable compared to even the improved Wasserstein GAN. Further, unlike the traditional GAN loss, the loss formulated in our method is a good measure of the actual distance between the distributions and, for the first time for GAN training, we are able to show estimates for the same.","Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2018-08-31 without embargo terms","The student, Ishan Deshpande, accepted the attached license on 2018-04-10 at 13:35.","The student, Ishan Deshpande, submitted this Thesis for approval on 2018-04-10 at 13:40.","This Thesis was approved for publication on 2018-04-11 at 14:56.","DSpace SAF Submission Ingestion Package generated from Vireo submission #12168 on 2018-08-31 at 17:11:38","Made available in DSpace on 2018-09-04T20:26:58Z (GMT). No. of bitstreams: 2 DESHPANDE-THESIS-2018.pdf: 14373491 bytes, checksum: beb5abc56a3c12e9d138b2f9708b96bb (MD5) LICENSE.txt: 4212 bytes, checksum: 210f1676a1810002500290549bf8041a (MD5) Previous issue date: 2018-04-11"],"dc:format":["application/pdf"],"dc:identifier":["http://hdl.handle.net/2142/100951"],"dc:language":["en"],"dc:rights":["Copyright 2018 Ishan Deshpande"],"dc:subject":["Machine Learning","Deep Learning"],"dc:title":["Generative modeling using the sliced Wasserstein distance"],"dc:type":["text"],"thesis:degree_discipline":["Electrical & Computer Engr"],"thesis:degree_level":["Thesis"],"thesis:degree_name":["M.S."],"thesis:institution_name":["University of Illinois at Urbana-Champaign"]},"updated_at":"2026-07-22T22:24:38Z"}