AI Paper Atlas

Research map / Federated learning and privacy / Personalized and heterogeneous FL

PRIOR: Personalized Prior for Reactivating the Information Overlooked in Federated Learning.

Mingjia Shi, Yuhao Zhou, Kai Wang, Huaizheng Zhang, Shudong Huang, Qing Ye, Jiancheng Lv

NeurIPS 2023 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Personalization with global-local trade-off · 5 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Combine loss-gradient and memorized envelope-gradient prior selection instead of using no personalized prior (ablation)pFedBreD · All tasks (Avg)accuracy (%): 77.6 -> 81.5Improved
Use loss-gradient prior selection alone instead of no personalized prior (ablation)pFedBreD · All tasks (Avg)accuracy (%): 77.6 -> 75.3Hurt
Use memorized envelope-gradient prior selection alone instead of no personalized prior (ablation)pFedBreD · Hard tasks (H.Avg)accuracy (%): 65.1 -> 65.7No clear effect
Add fine-tuning after the mh personalized modelpFedBreD mh · All tasks (Avg)accuracy (%): 81.47 -> 82.37Improved
Add aggregate momentum to the mh personalized modelpFedBreD mh · All tasks (Avg)accuracy (%): 81.47 -> 83.09Improved
Use mh+AM rather than the strongest listed baselineFEMNIST DNN · FEMNIST DNNaccuracy (%): 66.79 -> 70.34Improved
Use mh+AM rather than the strongest listed baselineSent140 LSTM · Sent140 LSTMaccuracy (%): 70.73 -> 73.68Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy