[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84243-en":3,"doc-seo-84243-105":29,"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":20,"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":13,"seo_description":14,"update_tm":27,"read_time":28},84243,13056703019662,"Evangeline","https://ap-avatar.wpscdn.com/avatar/be000253a8e92610077?_k=1778726343310543188",8,"Research & Report","Collaborative Synthetic Data Generation for Knowledge Transfer in Federated Learning","One-shot federated learning (OSFL) reduces communication by training in a single round, yet maintaining model quality under client data heterogeneity remains difficult. Prior approaches often transfer client knowledge via synthetic datasets or distillates, but typically lack rigorous privacy guarantees. FedKT-CSD introduces collaborative synthetic data generation using a shared latent space from publicly pretrained autoencoders, formal (ε, δ)-differential privacy, lightweight client computation, and scalable performance across heterogeneous datasets.","arXiv :2607 .07565v 1 [ cs .LG] 8 Jul 2026  \nCollaborative Synthetic Data Generation for Knowledge Transfer in Federated Learning  \n[Maximilian Andreas Hoefler](Maximilian Andreas Hoefler maximilian.andreas.hoefler@hhi.fraunhofer. de)[ maximilian.andreas.hoefler@hhi.fraunhofer. de](Maximilian Andreas Hoefler maximilian.andreas.hoefler@hhi.fraunhofer. de)  \nFraunhofer Heinrich Hertz Institute  \nKarsten Mueller [karsten.mueller@hhi.fraunhofer. de](karsten.mueller@hhi.fraunhofer. de)  \nFraunhofer Heinrich Hertz Institute  \nWojciech Samek [wojciech.samek@hhi.fraunhofer. de](wojciech.samek@hhi.fraunhofer. de)  \nFraunhofer Heinrich Hertz Institute BIFOLD  \nTechnical University Berlin  \nAbstract  \nOne-shot federated learning (OSFL) addresses the communication overhead of federated learning by limiting training to a single round, but doing so without sacrificing model quality is non-trivial, particularly when client data distributions diverge. Recent work has addressed this challenge by aggregating client knowledge on the server through the construction of transferable synthetic datasets or distillates. However, most of these methods lack formal privacy guarantees, leaving a gap in jointly achieving low communication, robustness to heterogeneity, and rigorous privacy. We propose FedKT-CSD (Federated Knowledge Transfer via Collaborative Synthetic Data), a framework inspired by neural image compression that closes this gap by leveraging publicly pretrained autoencoders as a shared latent space.  \nEach client encodes its private data in a single forward pass, computes class-conditional latent statistics, and transmits these to the server. The server aggregates these statistics via secure aggregation, adds calibrated differential privacy noise, and decodes a synthetic dataset for training a global model and further downstream tasks. This design provides formal (ε, δ)-differential privacy by construction, while keeping client-side computation and communication lightweight. Despite operating under privacy constraints, FedKT-CSD is competitive with and even outperforms non-private baselines across diverse datasets and heterogeneity settings, and scales to a large number of clients. Our code is available at:  \n[https://github.com/an7123/FedKT-CSD](https://github.com/an7123/FedKT-CSD)  \n1 Introduction  \nFederated Learning (FL) McMahan et al. (2017) has emerged as a powerful paradigm for training models collaboratively across decentralized data, under the promise of privacy preservation. Nonetheless, the presence of heterogeneous data across clients poses a significant challenge, often leading to slow convergence and suboptimal model performance. A wide range of approaches have been proposed to address this, from personalized federated learning (pFL) Luo & Wu (2022); Zhang et al. (2021); Dinh et al. (2020); Sun et al.(2021); Li et al. (2021); Collins et al. (2021) to data and representation sharing strategies Zhao et al. (2018); Zhu et al. (2021); Zhang et al. (2022b); Luo et al. (2024); Yang et al. (2023); Hoefler et al. (2025) . Despite their effectiveness, these methods typically require many communication rounds which can be problematic for real-world cross-device deployment.  \nA particularly attractive direction for communication constrained applications is one-shot federated learning (OSFL), where clients communicate with the server exactly once. Recent OSFL methods leverage the  \naforementioned data sharing principle, showing the efficacy in improving one-shot performance. Methods such as FedD3 Song et al. (2023), FedSD2C Zhang et al. (2024), FedCVAE Heinbaugh et al. (2023), DENSE Zhang et al. (2022a), CoBoosting Dai et al. (2024) and Beitollahi et al. (2025) approach OSFL by having each client train a local generative model or distill its data, then transmit the result to the server, which aggregates the contributions into a shared synthetic dataset or ensemble. These approaches thus eliminate repeated parameter exchanges, naturally sup","cbCaioqNsGh4YsXr","https://ap.wps.com/l/cbCaioqNsGh4YsXr","pdf",4400691,1,22,"English","en",105,"# Abstract\n# Introduction\n## Challenges in Federated and One-shot Federated Learning\n## Prior Knowledge-Transfer via Synthetic Data\n## Privacy Limitations and the Research Question\n## Proposed Approach with Shared Pretrained Autoencoders","[{\"question\":\"What problem does one-shot federated learning address, and why is it challenging?\",\"answer\":\"OSFL limits federated training to a single communication round to reduce overhead. Preserving model quality is challenging when clients have divergent, heterogeneous data distributions.\"},{\"question\":\"How does FedKT-CSD enable collaborative knowledge transfer while respecting privacy?\",\"answer\":\"Each client encodes private data into a shared latent space, computes class-conditional latent statistics, and sends them to the server. The server aggregates statistics using secure aggregation, adds calibrated differential privacy noise, and decodes a synthetic dataset for training the global model.\"},{\"question\":\"Why are publicly pretrained autoencoders central to FedKT-CSD’s design?\",\"answer\":\"They provide a shared latent representation without adaptation. Low-dimensional latent vectors have bounded sensitivity, enabling calibrated noise for formal (ε, δ)-differential privacy while keeping client-side computation and communication lightweight.\"}]",1784194308,55,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"collaborative-synthetic-data-generation-for-knowledge-transfer-in-federated-learning","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/collaborative-synthetic-data-generation-for-knowledge-transfer-in-federated-learning/84243/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does one-shot federated learning address, and why is it challenging?","Question",{"text":75,"@type":76},"OSFL limits federated training to a single communication round to reduce overhead. Preserving model quality is challenging when clients have divergent, heterogeneous data distributions.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does FedKT-CSD enable collaborative knowledge transfer while respecting privacy?",{"text":80,"@type":76},"Each client encodes private data into a shared latent space, computes class-conditional latent statistics, and sends them to the server. The server aggregates statistics using secure aggregation, adds calibrated differential privacy noise, and decodes a synthetic dataset for training the global model.",{"name":82,"@type":73,"acceptedAnswer":83},"Why are publicly pretrained autoencoders central to FedKT-CSD’s design?",{"text":84,"@type":76},"They provide a shared latent representation without adaptation. Low-dimensional latent vectors have bounded sensitivity, enabling calibrated noise for formal (ε, δ)-differential privacy while keeping client-side computation and communication lightweight.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"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":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]