Dual prompt personalized federated learning in foundation models
Abstract Personalized federated learning (PFL) has garnered significant attention for its ability to address heterogeneous client data distributions while preserving data privacy. However, when local client data is limited, deep learning models often suffer from insufficient training, leading to sub...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-07-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-025-11864-4 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849333272675352576 |
|---|---|
| author | Ying Chang Xiaohu Shi Xiaohui Zhao Zhaohuang Chen Deyin Ma |
| author_facet | Ying Chang Xiaohu Shi Xiaohui Zhao Zhaohuang Chen Deyin Ma |
| author_sort | Ying Chang |
| collection | DOAJ |
| description | Abstract Personalized federated learning (PFL) has garnered significant attention for its ability to address heterogeneous client data distributions while preserving data privacy. However, when local client data is limited, deep learning models often suffer from insufficient training, leading to suboptimal performance. Foundation models, such as CLIP (Contrastive Language-Image Pretraining), exhibit strong feature extraction capabilities and can alleviate this issue by fine-tuning on limited local data. Despite their potential, foundation models are rarely utilized in federated learning scenarios, and challenges related to integrating new clients remain largely unresolved. To address these challenges, we propose the Dual Prompt Personalized Federated Learning (DP2FL) framework, which introduces dual prompts and an adaptive aggregation strategy. DP2FL combines global task awareness with local data-driven insights, enabling local models to achieve effective generalization while remaining adaptable to specific data distributions. Moreover, DP2FL introduces a global model that enables prediction on new data sources and seamlessly integrates newly added clients without requiring retraining. Experimental results in highly heterogeneous environments validate the effectiveness of DP2FL’s prompt design and aggregation strategy, underscoring the advantages of prediction on novel data sources and demonstrating the seamless integration of new clients into the federated learning framework. |
| format | Article |
| id | doaj-art-21a9ff6e1d3445b6af80e963b3309d8f |
| institution | Kabale University |
| issn | 2045-2322 |
| language | English |
| publishDate | 2025-07-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj-art-21a9ff6e1d3445b6af80e963b3309d8f2025-08-20T03:45:56ZengNature PortfolioScientific Reports2045-23222025-07-0115111610.1038/s41598-025-11864-4Dual prompt personalized federated learning in foundation modelsYing Chang0Xiaohu Shi1Xiaohui Zhao2Zhaohuang Chen3Deyin Ma4College of Software, Jilin UniversityCollege of Software, Jilin UniversityCollege of Software, Jilin UniversityCollege of Computer Science and Technology, Jilin UniversityCollege of Computer Science and Engineering, Changchun University of TechnologyAbstract Personalized federated learning (PFL) has garnered significant attention for its ability to address heterogeneous client data distributions while preserving data privacy. However, when local client data is limited, deep learning models often suffer from insufficient training, leading to suboptimal performance. Foundation models, such as CLIP (Contrastive Language-Image Pretraining), exhibit strong feature extraction capabilities and can alleviate this issue by fine-tuning on limited local data. Despite their potential, foundation models are rarely utilized in federated learning scenarios, and challenges related to integrating new clients remain largely unresolved. To address these challenges, we propose the Dual Prompt Personalized Federated Learning (DP2FL) framework, which introduces dual prompts and an adaptive aggregation strategy. DP2FL combines global task awareness with local data-driven insights, enabling local models to achieve effective generalization while remaining adaptable to specific data distributions. Moreover, DP2FL introduces a global model that enables prediction on new data sources and seamlessly integrates newly added clients without requiring retraining. Experimental results in highly heterogeneous environments validate the effectiveness of DP2FL’s prompt design and aggregation strategy, underscoring the advantages of prediction on novel data sources and demonstrating the seamless integration of new clients into the federated learning framework.https://doi.org/10.1038/s41598-025-11864-4Personalized federated learningFoundation modelsClient heterogeneityAdaptive aggregation strategy |
| spellingShingle | Ying Chang Xiaohu Shi Xiaohui Zhao Zhaohuang Chen Deyin Ma Dual prompt personalized federated learning in foundation models Scientific Reports Personalized federated learning Foundation models Client heterogeneity Adaptive aggregation strategy |
| title | Dual prompt personalized federated learning in foundation models |
| title_full | Dual prompt personalized federated learning in foundation models |
| title_fullStr | Dual prompt personalized federated learning in foundation models |
| title_full_unstemmed | Dual prompt personalized federated learning in foundation models |
| title_short | Dual prompt personalized federated learning in foundation models |
| title_sort | dual prompt personalized federated learning in foundation models |
| topic | Personalized federated learning Foundation models Client heterogeneity Adaptive aggregation strategy |
| url | https://doi.org/10.1038/s41598-025-11864-4 |
| work_keys_str_mv | AT yingchang dualpromptpersonalizedfederatedlearninginfoundationmodels AT xiaohushi dualpromptpersonalizedfederatedlearninginfoundationmodels AT xiaohuizhao dualpromptpersonalizedfederatedlearninginfoundationmodels AT zhaohuangchen dualpromptpersonalizedfederatedlearninginfoundationmodels AT deyinma dualpromptpersonalizedfederatedlearninginfoundationmodels |