AI Paper Atlas

Research map / Federated learning and privacy / Communication-efficient FL

Sageflow: Robust Federated Learning against Both Stragglers and Adversaries

Jungwuk Park, Dong-Jun Han, Minseok Choi, Jaekyun Moon

NeurIPS 2021 · Federated learning and privacy › Communication-efficient FL · approach: Communication-efficient heterogeneous FL · 36 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
replace RFA with entropy filtering + loss-weighted averaging (ablation)Sageflow · model poisoning CIFAR10 (stragglers+adversaries)test accuracy: 10.00% -> 63.76%Improved
replace RFA with entropy filtering + loss-weighted averaging (ablation)Sageflow · scaled backdoor MNIST (stragglers+adversaries)attack success rate: 99.97% -> 0.21%Improved
use staleness-aware grouping instead of FedAsync for defenses (ablation)Sageflow · model poisoning FMNIST (stragglers+adversaries)test accuracy: 66.38% -> 85.23%Improved
use Sageflow instead of Zeno+Sageflow · model poisoning MNISTtest accuracy: 24.09% -> 97.55%Improved
use Sageflow instead of RFASageflow · model poisoning CIFAR10test accuracy: 16.43% -> 65.47%Improved
use Sageflow instead of RFA on data poisoning CIFAR10Sageflow · data poisoning CIFAR10test accuracy: 65.02% -> 64.46%Hurt
use Sageflow instead of FedAvgSageflow · scaled backdoor FMNIST/MNISTattack success rate: 99.99% -> 0.39%Improved
use Sageflow instead of ignore-stragglersSageflow · with only stragglersaccuracy vs timeImproved
Sageflow vs FedAsync with only stragglersSageflow · MNIST/FMNIST/CIFAR10 stragglers onlyaccuracy vs timeNo clear effect
remove entropy filtering or loss-weighted averaging alone (ablation)Sageflow · adversary attackstest accuracyHurt

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