AI Paper Atlas

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

CLoVE: Personalized Federated Learning through Clustering of Loss Vector Embeddings

Randeep Bhatia, Nikos Papadis, Murali Kodialam, T. V. Lakshman, Sayak Chakrabarty

ICML 2026 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Bayesian personalized FL

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 IFCA's lowest-loss assignment with clustering of loss vectorsIFCA · CIFAR-10, label skew 1accuracy: 85.7 -> 90.5Improved
replace IFCA's lowest-loss assignment with clustering of loss vectorsIFCA · MNIST, concept shiftaccuracy: 80.0 -> 97.6Improved
replace IFCA's lowest-loss assignment with clustering of loss vectorsIFCA · CIFAR-10, concept shiftaccuracy: 52.7 -> 58.4Improved
replace IFCA's lowest-loss assignment with clustering of loss vectorsIFCA · MNIST, label skew 1ARI: 0.83 -> 1.00Improved
use loss-vector embeddings instead of FlexCFL's one-shot gradient clusteringFlexCFL · CIFAR-10, concept shiftARI: 0.22 -> 1.00Improved
cluster clients by loss vectors instead of using a single global FedAvg modelFedAvg · CIFAR-10, feature skewaccuracy: 52.4 -> 57.1Improved
cluster clients instead of training a single global FedAvg modelFedAvg · Amazon Reviewsaccuracy: 99.0 -> 86.8Hurt
use CLoVE instead of FlexCFLFlexCFL · AG News, label skew 2accuracy: 60.2 -> 55.1Hurt
use CLoVE instead of IFCA for autoencoder trainingIFCA · MNIST (unsupervised)reconstruction test loss: 0.024 -> 0.014Improved
use CLoVE instead of IFCA for autoencoder trainingIFCA · CIFAR-10 (unsupervised)reconstruction test loss: 0.019 -> 0.021Hurt
use CLoVE's loss-vector clustering instead of CFL-S gradient clustering for fast cluster recoveryCFL-S · MNIST, concept shiftfirst round with ARI >= 0.9: 60.0 -> 2.0Improved
vary model initialization (same vs different) and first-round assignment (random vs evaluation-based) (ablation)CLoVE and IFCA · MNIST unsupervised, 10 clustersclustering accuracy over timeImproved

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