Filtering-based robust multi-task gradient descent matches minimax rates under task contamination and heterogeneity, removing the √d contamination barrier of regularization and score-based methods.
arXiv preprint arXiv:2602.20698 , year =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2026 1verdicts
ACCEPT 1representative citing papers
citing papers explorer
-
Contaminated Multi-task Learning with Heterogeneity: Fundamental Limits and Optimal Algorithms
Filtering-based robust multi-task gradient descent matches minimax rates under task contamination and heterogeneity, removing the √d contamination barrier of regularization and score-based methods.