[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-119350-en":3,"doc-seo-119350-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},119350,1099514067415,"Rowan","https://ap-avatar.wpscdn.com/avatar/100002539d78ffe74a7?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779092875211072502",8,"Research & Report","Privacy-Aware Synthetic Data Generation and Knowledge Transfer for Machine Learning","Privacy-Aware Synthetic Data Generation and Knowledge Transfer for Machine Learning addresses privacy and regulatory constraints that limit direct data sharing across institutions, especially for sensitive multimedia such as medical images and behavioral video. The work develops differentially private generative approaches and knowledge transfer strategies to enable safer publishing and adaptation of models. It covers methods including differentially private GANs with model inversion, Gaussian-optimized data publishing, latent-space privacy mechanisms, and MLLM-driven source-free curriculum learning for improved knowledge transfer.","UC Davis  \nUC Davis Electronic Theses and Dissertations  \nTitle  \nPrivacy-Aware Synthetic Data Generation and Knowledge Transfer for Machine Learning  \nPermalink  \n[https://escholarship.org/uc/item/30k015ct](https://escholarship.org/uc/item/30k015ct)  \nAuthor  \nChen, Dongjie  \nPublication Date  \n2024  \nPeer reviewed|Thesis/dissertation  \n[eScholarship.org](eScholarship.org) Powered by the California Digital Library  \nUniversity of California  \nPrivacy-Aware Synthetic Data Generation and Knowledge Transfer for Machine Learning  \nBy  \nDongjie Chen  \nDissertation  \nSubmitted in partial satisfaction of the requirements for the degree of  \nDoctor of Philosophy  \nin  \nElectrical and Computer Engineering in the  \nOFFICE OF GRADUATE STUDIES of the  \nUniversity of California  \nDavis  \nApproved:  \n\n| Chen-Nee Chuah, Chair |\n| --- |\n| Sen-Ching Samson Cheung, Co-Chair |\n| Sally Ozonoff |\n\nSoheil Ghiasi Committee in Charge 2024  \nCopyright © 2024 by Dongjie Chen All rights reserved.  \nContents  \nAbstract ....................................... v  \n[Acknowledgments .................................. vi](Acknowledgments .................................. vi)  \n1 Introduction 1  \n1.1 Motivations .................................. 2  \n1.2 Contributions ................................. 7  \n1.2.1 DPMI: Differentially Private Generative Adversarial Networks with Model Inversion ............................ 7  \n1.2.2 DPGOMI: Differentially Private Data Publishing with Gaussian Optimized Model Inversion ...................... 8  \n1.2.3 DP-PAMI: A Latent Space Solution for Differentially Private Synthetic Data Generation ........................ 8  \n1.2.4 RCL: Empowering Source-Free Domain Adaptation with MLLMdriven Curriculum Learning ..................... 9  \n2 Backgrounds 10  \n2.1 Generative Adversarial Network (GAN) ................... 10  \n2.1.1 GAN and its Variants ........................ 10  \n2.1.2 GAN Inversion ............................ 11  \n2.2 Differential Privacy .............................. 12  \n2.3 Differentially Private GANs ......................... 13  \n2.3.1 RDP Accountant ........................... 15  \n2.4 Domain Adaptation .............................. 16  \n2.4.1 Unsupervised Domain Adaptation .................. 16  \n2.4.2 Source-Free Domain Adaptation ................... 16  \n2.5 Vision-Language Models ........................... 18  \n2.6 Multi-modal Large Language Models (MLLM) ............... 18  \n3 DPMI: Differentially Private Generative Adversarial Networks with Model Inversion 22  \n3.1 Introduction .................................. 22  \n3.2 Methodology ................................. 26  \n3.3 Proposed DPMI Framework ......................... 27  \n3.3.1 Model Inversion of GANs ...................... 27  \n3.3.2 Differentially private GAN on the latent space ........... 29  \n3.4 Experiments .................................. 29  \n3.4.1 Dataset Partition ........................... 30  \n3.4.2 Datasets and Evaluation Metrics .................. 30  \n3.4.3 Ablation study on Model Inversion ................. 31  \n3.4.4 Comparison with DP-GAN ...................... 33  \n4 DPGOMI: Differentially Private Data Publishing with Gaussian Optimized Model Inversion 37  \n4.1 Introduction .................................. 37  \n4.2 Related Work ................................. 38  \n4.3 Proposed Method ............................... 39  \n4.4 Experiments .................................. 40  \n4.4.1 Dataset Partition ........................... 40  \n4.4.2 Datasets and Evaluation Metrics .................. 42  \n4.4.3 Comparison with SOTA DP generative models ........... 43  \n4.4.4 Ablation study on Gaussian Optimized Model Inversion ...... 44  \n5 DP-PAMI: A Latent Space Solution for Differentially Private Synthetic Data Generation 46  \n5.1 Introduction .................................. 46  \n5.2 Related Work ................................. 48  \n5.3 Preliminaries ................................. 49  \n5.4 Proposed Method ........","cbCailQuJgXQqvOs","https://ap.wps.com/l/cbCailQuJgXQqvOs","pdf",16215224,1,112,"English","en",105,"# Abstract\n# Introduction\n## Motivations\n## Contributions\n## DPMI: Differentially Private Generative Adversarial Networks with Model Inversion\n## DPGOMI: Differentially Private Data Publishing with Gaussian Optimized Model Inversion\n## DP-PAMI: A Latent Space Solution for Differentially Private Synthetic Data Generation\n## RCL: Empowering Source-Free Domain Adaptation with MLLM-driven Curriculum Learning\n# Backgrounds\n## Generative Adversarial Network (GAN)\n## Differential Privacy\n## Differentially Private GANs\n## Domain Adaptation\n## Vision-Language Models\n## Multi-modal Large Language Models (MLLM)\n# Conclusions and Future Works","[{\"question\":\"Why does the thesis focus on privacy-aware synthetic data for machine learning?\",\"answer\":\"It targets the gap between the need for diverse labeled data to improve generalization and the privacy/regulatory restrictions that often prevent sharing sensitive multimedia data.\"},{\"question\":\"What is DPMI in the proposed work?\",\"answer\":\"DPMI is a differentially private generative adversarial approach that combines generative modeling with model inversion to enable privacy-aware learning and data-related transfer.\"},{\"question\":\"How does RCL support source-free domain adaptation?\",\"answer\":\"RCL uses MLLM-driven curriculum learning with pseudo-labeling, reliability measurement, and knowledge expansion to transfer knowledge without relying on the original source.\"}]","Privacy-Aware Synthetic Data Generation and Knowledge Transfer for Machine Learning | PDF",1785723850,282,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"privacy-aware-synthetic-data-generation-and-knowledge-transfer-for-machine-learning","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/privacy-aware-synthetic-data-generation-and-knowledge-transfer-for-machine-learning/119350/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why does the thesis focus on privacy-aware synthetic data for machine learning?","Question",{"text":75,"@type":76},"It targets the gap between the need for diverse labeled data to improve generalization and the privacy/regulatory restrictions that often prevent sharing sensitive multimedia data.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is DPMI in the proposed work?",{"text":80,"@type":76},"DPMI is a differentially private generative adversarial approach that combines generative modeling with model inversion to enable privacy-aware learning and data-related transfer.",{"name":82,"@type":73,"acceptedAnswer":83},"How does RCL support source-free domain adaptation?",{"text":84,"@type":76},"RCL uses MLLM-driven curriculum learning with pseudo-labeling, reliability measurement, and knowledge expansion to transfer knowledge without relying on the original source.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]