AI Paper Atlas

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

FLuID: Mitigating Stragglers in Federated Learning using Invariant Dropout

Irene Wang, Prashant J. Nair, Divya Mahajan

NeurIPS 2023 · Federated learning and privacy › Communication-efficient FL · approach: Efficient local training in FL · 3 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 invariant-neuron dropout instead of Ordered DropoutFederated learning with stragglers · Shakespeare, CIFAR10, FEMNISTaccuracy: maximum increase of 1.4%Improved
Use invariant-neuron dropout instead of Random DropoutFederated learning with stragglers · Shakespeare, CIFAR10, FEMNISTaccuracy: maximum accuracy gain of 1.6% pointsImproved
Dynamically recalibrate stragglers and sub-model sizes at runtimeFLuID · Mobile-client training with changing stragglerstraining time: 18% to 26% faster training time compared to the baselineImproved
Use dynamic rather than static straggler selectionFLuID · Mobile-client training with changing stragglerstraining time: 14% to 18% faster training timeImproved
Increase the invariant-neuron threshold from 1% to 8% (ablation)Invariant Dropout · FEMNIST, sub-model size 0.75accuracy: 80.1 -> 80.7Improved
Increase the invariant-neuron threshold from 8% to 10% (ablation)Invariant Dropout · FEMNIST, sub-model size 0.75accuracy: 80.7 -> 80.5Hurt
Cluster stragglers by capability and assign multiple sub-model sizesFLuID · CIFAR10, FEMNIST, Shakespeareaccuracy: Invariant: 72.7, 78.2, 54.1; Ordered: 72.3, 77.4, 53.9Improved
Train stragglers on a 0.75-sized sub-model as straggler percentage rises (ablation)Invariant Dropout · Shakespeare, CIFAR10, FEMNISTaccuracyHurt

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