AI Paper Atlas

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

Flow: Per-instance Personalized Federated Learning

Kunjal Panchal, Sunav Choudhary, Nisarg Parikh, Lijun Zhang, Hui Guan

NeurIPS 2023 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Personalization with global-local trade-off · 2 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
remove regularization term from policy update (ablation)Flow · Stackoverflowgeneralized accuracy: 24.16% -> 26.64%Improved
remove regularization term from policy update (ablation)Flow · Stackoverflowpersonalized accuracy: 27.59% -> 29.49%Improved
per-instance routing instead of both paths being global (PI-Flow) (ablation)Flow · Stackoverflowpersonalized accuracy: 26.31% -> 29.49%Improved
per-instance and per-client Flow instead of per-client only (PC-Flow, FedAvgFT) (ablation)Flow · Stackoverflowpersonalized accuracy: 24.41% -> 29.49%Improved
dynamic routing instead of fixed instance-agnostic policy q0 (ablation)Flow · Stackoverflowvalidation accuracy (personalized)Improved
fixed policy q0 = 0.25 (only local route at inference) (ablation)Flow · Stackoverflowvalidation accuracy (personalized)Hurt
soft decisions instead of hard decisions at inference (ablation)Flow · Stackoverflowpersonalized accuracy: 29.49% -> 29.57%No clear effect
Flow instead of best baselineFlow · Stackoverflow, Shakespeare, EMNIST, CIFAR10/100generalized accuracy: +1.11-3.46%Improved
Flow instead of best baselineFlow · Stackoverflow, Shakespeare, EMNIST, CIFAR10/100personalized accuracy: +1.33-4.58%Improved
Flow instead of KNNPER (other per-instance method)Flow · Stackoverflow, Shakespeare, CIFAR10personalized accuracy: +1.33-6.97%Improved
Flow instead of PartialFedFlow · CIFAR100 (0.1)generalized accuracy: 34.79% -> 34.00%Hurt
Flow instead of PartialFedFlow · CIFAR100 (0.6)personalized accuracy: 40.18% -> 40.08%No 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