AI Paper Atlas

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

FRAug: Tackling Federated Learning with Non-IID Features via Representation Augmentation

Haokun Chen, Ahmed Frikha, Denis Krompass, Jindong Gu, Volker Tresp

ICCV 2023 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Federated learning with pretrained models · 50 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 FRAug representation augmentation to FedAvg-style trainingFedAvg · OfficeHome (avg)accuracy: 64.69 -> 66.60Improved
add FRAug to FedAvgFedAvg · Digits (avg)accuracy: 85.62 -> 89.59Improved
add a shared generator G to the client-specific RTNet (ablation)FRAug · OfficeHome (avg)accuracy: 64.58 -> 65.99Improved
add RTNet transformation to client-agnostic embeddings (ablation)FRAug · Digits (avg)accuracy: 87.25 -> 89.51Improved
add RTNet to generator on PACS (ablation)FRAug · PACS (avg)accuracy: 90.82 -> 91.35Improved
add EMA class-prototype embeddings (ablation)FRAug · OfficeHome (avg)accuracy: 65.99 -> 66.60Improved
add EMA class-prototype embeddings (ablation)FRAug · PACS (avg)accuracy: 91.35 -> 91.05No clear effect
add additional real-image embeddings to finetune prediction headFedAvg classifier · OfficeHome (motivation)accuracy: 64.69 -> 75.01Improved
replace random-noise embedding augmentation with learned FRAug augmentationprediction head training · OfficeHome (avg)accuracy: 65.06 -> 66.60Improved
apply input-space augmentation methods FAug and FedReg instead of FedAvgFedAvg · OfficeHome (avg)accuracy: 64.69 -> 60.39 (FAug), 62.00 (FedReg)Hurt
scale to more clients on RxRx1FRAug · RxRx1 (avg)accuracy: 30.84 -> 32.07 -> 33.20 (8, 12, 16 clients)Improved
add FRAug under strong input noiseFedAvg · Noisy OfficeHome (sigma=1.0)accuracy: 31.26 -> 31.65No clear effect

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