A two-stage federated learning framework detects noisy-label clients, corrects their labels via masked learnable distributions, and aggregates with geometric median weights.
Tsang, and Masashi Sugiyama
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Robust Federated Learning against Noisy Clients via Masked Optimization
A two-stage federated learning framework detects noisy-label clients, corrects their labels via masked learnable distributions, and aggregates with geometric median weights.