CAB: Continuous Adaptive Blending for Policy Evaluation and Learning

By Yi-Hsun Su, Lequn Wang, Michele Santacatterina, Thorsten Joachims. 2019

91 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Noveen Sachdeva
2. Yi-Hsun Su
3. Yuta Saito
4. Suguru Yaginuma
5. Yuta Nishino
6. Hayato Sakata
7. Kazuhide Nakata
8. Thorsten Joachims

Citations in the corpus

• Off-policy Bandits with Deficient Support (2020, methodology): “To this effect, virtually all state-of-the-art off-policy learning methods for contextual-bandit problems rely on counterfactual estimators [3, 5, 6, 22, 24, 26] that employ inverse propensity score (IPS) weighting to get an unbiased ERM objective.”
• Unbiased Recommender Learning from Missing-Not-At-Random Implicit Feedback (2019, methodology): “the true loss in the implicit feedback setting. However, Theorem 4.4 suggests that the unbiased estimator can suffer from high variance. In this subsection, we apply the propensity clipping technique [3, 5, 25] to our estimator and analyze the bias-variance trade-off of an estimator with the clippin”