A doubly robust estimator and a policy learner achieve minimax optimal regret κ(Π)/√n under concept drift with only conditional Y|X shifts.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Distributionally Robust Policy Learning under Concept Drifts
A doubly robust estimator and a policy learner achieve minimax optimal regret κ(Π)/√n under concept drift with only conditional Y|X shifts.