AI Paper Atlas

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

Global Balanced Experts for Federated Long-Tailed Learning

Yaopei Zeng, Lei Liu, Li Liu, Li Shen, Shaoguo Liu, Baoyuan Wu

ICCV 2023 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Federated learning with pretrained models · 21 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
use global label distribution instead of local label distribution as class prior for BSM (ablation)BSM · CIFAR-100-LT IR=100top-1 accuracy: 35.55 -> 38.56Improved
use GPI as class prior instead of local label distribution (ablation)BSM · CIFAR-100-LT IR=100top-1 accuracy: 35.55 -> 37.19Improved
add multi-expert ensemble on top of GPI prior (ablation)BSM with GPI · CIFAR-100-LT IR=100top-1 accuracy: 37.19 -> 40.42Improved
group clients randomly instead of by LPI/GPI similarity (ablation)GBME · CIFAR-100-LT IR=100top-1 accuracy: 37.19 -> 37.38No clear effect
group clients by GPI cosine similarity (ablation)GBME · CIFAR-100-LT IR=100top-1 accuracy: 37.19 -> 40.42Improved
increase number of experts from 1 to 4 (ablation)GBME · CIFAR-100-LT IR=100top-1 accuracy: 37.19 -> 41.68Improved
increase experts from 4 to 5 (ablation)GBME · CIFAR-100-LT IR=100top-1 accuracy: 41.68 -> 41.85No clear effect
client selection alpha at 0 or 1 versus about 0.6 (multiple selection) (ablation)GBME · CIFAR-100-LT IR=100top-1 accuracy: 36.19 -> 40.42Improved
tighten privacy budget (lower epsilon) in GBME-p (ablation)GBME-p · CIFAR-100-LT IR=100top-1 accuracy: 40.24 -> 38.50Hurt
use global instead of local label prior for re-balance methods (RIDE) (ablation)RIDE · CIFAR-100-LT IR=100top-1 accuracy: 34.94 -> 34.59No clear effect
GBME with half communication rounds versus FedAvgGBME · CIFAR-100-LT IR=100top-1 accuracy: 34.48 -> 35.11Improved

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