AI Paper Atlas

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

FedPerfix: Towards Partial Model Personalization of Vision Transformers in Federated Learning

Guangyu Sun, Matias Mendieta, Jun Luo, Shandong Wu, Chen Chen

ICCV 2023 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Sparse and collaboration-graph personalization · 31 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 adapter-generated prefixes (FedPerfix) instead of vanilla attention personalizationViT-Small PFL · CIFAR-100Top-1 accuracy: 44.63 -> 48.10Improved
generate prefixes with parallel adapter instead of zero-initialized prefixes (ablation)FedPerfix · CIFAR-100Top-1 accuracy: 47.37 -> 48.10Improved
generate prefixes with parallel adapter instead of random-init prefixes (ablation)FedPerfix · CIFAR-100Top-1 accuracy: 46.98 -> 48.10Improved
add personalized prompts on inputs (ablation)ViT-Small PFL · CIFAR-100Top-1 accuracy: 44.19 vs 44.42 (head local)No clear effect
use MLP adapters instead of prefixes (ablation)FedPerfix · CIFAR-100Top-1 accuracy: 48.10 -> 47.99No clear effect
keep self-attention layers local alongside head (ablation)ViT-Small FedAvg · CIFAR-100Top-1 accuracy: 23.29 -> 42.95Improved
keep LayerNorm local (FedBN-style) (ablation)ViT-Small FedAvg · CIFAR-100Top-1 accuracy: 23.29 -> 23.94No clear effect
replace ResNet50 with ViT-Small backbone (ablation)FedRep and other PFL methods · CIFAR-100Top-1 accuracy: FedRep 39.52 -> 44.42Improved
increase model size from ViT-Small to ViT-Base (ablation)FedPerfix · CIFAR-100Top-1 accuracy: 48.10 -> 48.40Improved
lower client participation rate to 6.25% (ablation)FedPerfix · CIFAR-100 (N=64)Top-1 accuracy: 48.10 -> 43.80Hurt
Per-FedAVG meta-learning on ViTFedAvg · Office-HomeTop-1 accuracy: 21.47 -> 17.09Hurt

Later papers that compared with this method

1 later accepted paper reports FedPerfix as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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