AI Paper Atlas

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

Personalized Federated Domain-Incremental Learning based on Adaptive Knowledge Matching

Yichen Li, Wenchao Xu, Haozhao Wang, Yining Qi, Jingcai Guo, Ruixuan Li

ECCV 2024 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Non-IID federated learning · 17 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
add knowledge migration loss (ablation)pFedDIL · Digit-10average accuracy (%): 86.45 -> 89.28Improved
add knowledge migration loss (ablation)pFedDIL · DomainNetaverage accuracy (%): 48.79 -> 51.69Improved
add correlation-weighted ensemble inference (ablation)pFedDIL · Digit-10average accuracy (%): 87.73 -> 89.28Improved
add partial parameter sharing between auxiliary classifier and model (ablation)pFedDIL · Digit-10average accuracy (%): 89.18 -> 89.28No clear effect
add partial parameter sharing (ablation)pFedDIL · Office-31average accuracy (%): 45.52 -> 46.17No clear effect
use pFedDIL versus FedCILfederated domain-incremental learning · DomainNetaverage accuracy (%): 47.64 -> 51.69Improved
use pFedDIL versus FedAvg in communication costfederated learning · Digit-10rounds to best performance: 480 -> 379Improved
use pFedDIL versus FedAvg in time per roundfederated learning · Digit-10seconds per round: 32.14 -> 34.91Hurt
raise lambda from 0.0 to 1.0 (more personalized models) (ablation)pFedDIL · Digit-10average accuracy (%) / model size: 82.43 -> 93.94; 22Mb -> 88MbMixed
use disjoint per-domain models (upper bound) versus pFedDILpFedDIL · DomainNetaverage accuracy (%): 51.69 -> 53.26Improved

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