Each local estimator of a decentralized robust kernel-based gradient descent achieves optimal regression rates (up to logs) when local data sizes and the robustness scale are chosen according to explicit conditions.
Duchi, Alekh Agarwal, Martin J
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
-
Theory of Decentralized Robust Kernel-Based Learning
Each local estimator of a decentralized robust kernel-based gradient descent achieves optimal regression rates (up to logs) when local data sizes and the robustness scale are chosen according to explicit conditions.