UDJ-FL is a single federated learning objective whose hyperparameters select among four distributive-justice fairness notions, with client weights set by aleatoric uncertainty.
Title resolution pending
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
-
Achieving Distributive Justice in Federated Learning via Uncertainty Quantification
UDJ-FL is a single federated learning objective whose hyperparameters select among four distributive-justice fairness notions, with client weights set by aleatoric uncertainty.