{"id":{"repo_id":"gatech","oai_identifier":"oai:repository.gatech.edu:1853/73167"},"canonical_url":"https://search.ndltd.org/etd/gatech/oai:repository.gatech.edu:1853/73167","repository":{"repo_id":"gatech","name":"Georgia Tech","base_url":"https://repository.gatech.edu/server/oai/request"},"display":{"title":"Controllability and Uncertainty in Generative Models","abstract":"This dissertation describes methods for enhancing generative models with either added controllability or expressiveness of uncertainty, demonstrating how a strong prior enables both features. One general approach is to introduce new architectures or training objectives. However, current trends towards massive upscaling of model size, training data, and computational resources can make retraining or fine-tuning difficult and expensive. Thus, another approach is to build upon existing pre-trained models. We consider both types of approaches with an emphasis on the latter. We first tackle the tasks of controllable image synthesis and uncertainty estimation through training-based methods and then switch focus towards computationally-efficient methods that do not require direct updates to the base model's parameters. We conclude by discussing future directions based on the insights from our findings.","abstract_html":"This dissertation describes methods for enhancing generative models with either added controllability or expressiveness of uncertainty, demonstrating how a strong prior enables both features. One general approach is to introduce new architectures or training objectives. However, current trends towards massive upscaling of model size, training data, and computational resources can make retraining or fine-tuning difficult and expensive. Thus, another approach is to build upon existing pre-trained models. We consider both types of approaches with an emphasis on the latter. We first tackle the tasks of controllable image synthesis and uncertainty estimation through training-based methods and then switch focus towards computationally-efficient methods that do not require direct updates to the base model&#x27;s parameters. We conclude by discussing future directions based on the insights from our findings.","abstract_has_math":false,"creators":["Ham, Cusuh"],"institution":"Georgia Institute of Technology","degree_name":null,"degree_level":"Doctoral","degree_discipline":null,"degree_department":"Interactive Computing","school":null,"contributors":[],"advisors":["Hays, James"],"committee_chairs":[],"committee_members":["Hoffman, Judy","Kira, Zsolt","Shi, Humphrey","Zhu, Jun-Yan"],"year":2023,"date_issued":"2023-12-06","date_published":"2023-12-06","updated_at":"2026-11-19T06:03:21Z","subjects":["Generative models","Image synthesis","Multimodal learning","Conditional synthesis","Uncertainty estimation"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[],"person_identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/1853/73167","outbound_label":"Handle","outbound_source":"dc:identifier.uri","record_url":"https://hdl.handle.net/1853/73167","record_label":"Repository record","record_kind":"handle","record_source":"dc:identifier.uri","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://repository.gatech.edu/server/oai/request?verb=GetRecord&metadataPrefix=dim&identifier=oai%3Arepository.gatech.edu%3A1853%2F73167","prefix":"dim"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Hays, James"]},{"key":"dc:contributor.committeemember","label":"Committee Member","values":["Hoffman, Judy","Kira, Zsolt","Shi, Humphrey","Zhu, Jun-Yan"]},{"key":"dc:contributor.department","label":"Department","values":["Interactive Computing"]},{"key":"dc:contributor.other","label":"Dc Contributor Other","values":["College of Computing","School of Interactive Computing"]},{"key":"dc:creator","label":"Author","values":["Ham, Cusuh"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2024-01-10T18:50:03Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2024-01-10T18:50:03Z"]},{"key":"dc:date.created","label":"Dc Date Created","values":["2023-12"]},{"key":"dc:date.issued","label":"Date","values":["2023-12-06"]},{"key":"dc:date.submitted","label":"Dc Date Submitted","values":["December 2023"]},{"key":"dc:date.updated","label":"Dc Date Updated","values":["2024-01-10T18:50:03Z"]},{"key":"dc:publisher","label":"Institution","values":["Georgia Institute of Technology"]},{"key":"dc:relation.advisorentity","label":"Dc Relation Advisorentity","values":["5b631b9d-32a1-48bb-887b-c5431096e10d"]},{"key":"dc:relation.orgunit","label":"Dc Relation Orgunit","values":["c8892b3c-8db6-4b7b-a33a-1b67f7db2021","aac3f010-e629-4d08-8276-81143eeaf5cc"]},{"key":"dc:type","label":"Dc Type","values":["Text","Dissertation"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Doctoral"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Generative models","Image synthesis","Multimodal learning","Conditional synthesis","Uncertainty estimation"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/1853/73167"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["This dissertation describes methods for enhancing generative models with either added controllability or expressiveness of uncertainty, demonstrating how a strong prior enables both features. One general approach is to introduce new architectures or training objectives. However, current trends towards massive upscaling of model size, training data, and computational resources can make retraining or fine-tuning difficult and expensive. Thus, another approach is to build upon existing pre-trained models. We consider both types of approaches with an emphasis on the latter. We first tackle the tasks of controllable image synthesis and uncertainty estimation through training-based methods and then switch focus towards computationally-efficient methods that do not require direct updates to the base model's parameters. We conclude by discussing future directions based on the insights from our findings."]},{"key":"dc:description.degree","label":"Dc Description Degree","values":["Ph.D."]},{"key":"dc:format.mimetype","label":"Dc Format Mimetype","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Controllability and Uncertainty in Generative Models"]},{"key":"dspace:entity.type","label":"Dspace Entity Type","values":["Publication"]}]}],"canonical_facts":{"dc:contributor.advisor":["Hays, James"],"dc:contributor.committeemember":["Hoffman, Judy","Kira, Zsolt","Shi, Humphrey","Zhu, Jun-Yan"],"dc:contributor.department":["Interactive Computing"],"dc:contributor.other":["College of Computing","School of Interactive Computing"],"dc:creator":["Ham, Cusuh"],"dc:date.accessioned":["2024-01-10T18:50:03Z"],"dc:date.available":["2024-01-10T18:50:03Z"],"dc:date.created":["2023-12"],"dc:date.issued":["2023-12-06"],"dc:date.submitted":["December 2023"],"dc:date.updated":["2024-01-10T18:50:03Z"],"dc:description.abstract":["This dissertation describes methods for enhancing generative models with either added controllability or expressiveness of uncertainty, demonstrating how a strong prior enables both features. One general approach is to introduce new architectures or training objectives. However, current trends towards massive upscaling of model size, training data, and computational resources can make retraining or fine-tuning difficult and expensive. Thus, another approach is to build upon existing pre-trained models. We consider both types of approaches with an emphasis on the latter. We first tackle the tasks of controllable image synthesis and uncertainty estimation through training-based methods and then switch focus towards computationally-efficient methods that do not require direct updates to the base model's parameters. We conclude by discussing future directions based on the insights from our findings."],"dc:description.degree":["Ph.D."],"dc:format.mimetype":["application/pdf"],"dc:identifier.uri":["https://hdl.handle.net/1853/73167"],"dc:publisher":["Georgia Institute of Technology"],"dc:relation.advisorentity":["5b631b9d-32a1-48bb-887b-c5431096e10d"],"dc:relation.orgunit":["c8892b3c-8db6-4b7b-a33a-1b67f7db2021","aac3f010-e629-4d08-8276-81143eeaf5cc"],"dc:subject":["Generative models","Image synthesis","Multimodal learning","Conditional synthesis","Uncertainty estimation"],"dc:title":["Controllability and Uncertainty in Generative Models"],"dc:type":["Text","Dissertation"],"dspace:entity.type":["Publication"],"thesis:degree_level":["Doctoral"]},"updated_at":"2026-11-19T06:03:21Z"}