CSI converts offline contextual bandit learning into a binary classification problem by comparing the logged action against a counterfactual action sampled from the logging policy, and the argmax of the resulting classifier provably matches the argmax of expected reward.
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
-
Offline Contextual Bandit with Counterfactual Sample Identification
CSI converts offline contextual bandit learning into a binary classification problem by comparing the logged action against a counterfactual action sampled from the logging policy, and the argmax of the resulting classifier provably matches the argmax of expected reward.