Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Provably Personalized and Robust Federated Learning

About

Identifying clients with similar objectives and learning a model-per-cluster is an intuitive and interpretable approach to personalization in federated learning. However, doing so with provable and optimal guarantees has remained an open challenge. We formalize this problem as a stochastic optimization problem, achieving optimal convergence rates for a large class of loss functions. We propose simple iterative algorithms which identify clusters of similar clients and train a personalized model-per-cluster, using local client gradients and flexible constraints on the clusters. The convergence rates of our algorithms asymptotically match those obtained if we knew the true underlying clustering of the clients and are provably robust in the Byzantine setting where some fraction of the clients are malicious.

Mariel Werner, Lie He, Michael Jordan, Martin Jaggi, Sai Praneeth Karimireddy• 2023

Related benchmarks

TaskDatasetResultRank
Image ClassificationMNIST-Rotation (test)
Avg Acc84.23
16
Image ClassificationMNIST Label-Switching (test)
Accuracy (Test)92.18
8
Image ClassificationCIFAR-100 sigma=10 (test)
Test Accuracy28.77
8
Image ClassificationCIFAR-100 sigma=0.01 (test)
Test Accuracy24.99
8
Next-Character PredictionShakespeare n=15, m=7 (test)
Test Accuracy53.4
7
Next-Character PredictionShakespeare (n=30, m=3) (test)
Accuracy48.73
7
Showing 6 of 6 rows

Other info

Follow for update