Doubly robust off-policy evaluation with shrinkage
arXiv:1907.09623
Abstract
We propose a new framework for designing estimators for off-policy evaluation in contextual bandits. Our approach is based on the asymptotically optimal doubly robust estimator, but we shrink the importance weights to minimize a bound on the mean squared error, which results in a better bias-variance tradeoff in finite samples. We use this optimization-based framework to obtain three estimators: (a) a weight-clipping estimator, (b) a new weight-shrinkage estimator, and (c) the first shrinkage-based estimator for combinatorial action sets. Extensive experiments in both standard and combinatorial bandit benchmark problems show that our estimators are highly adaptive and typically outperform state-of-the-art methods.
References in corpus (7)
- Unbiased Offline Evaluation of Contextual-bandit-based News Article Recommendation Algorithms
- Doubly Robust Policy Evaluation and Learning
- Counterfactual Risk Minimization: Learning from Logged Bandit Feedback
- Learning from Logged Implicit Exploration Data
- Balanced Policy Evaluation and Learning
- A Framework for Optimal Matching for Causal Inference
- CAB: Continuous Adaptive Blending Estimator for Policy Evaluation and Learning
Cited by in corpus (6)
- A Survey on Causal Inference for Recommendation
- Doubly Robust Off-Policy Evaluation for Ranking Policies under the Cascade Behavior Model
- Evaluating the Robustness of Off-Policy Evaluation
- Ad-load Balancing via Off-policy Learning in a Content Marketplace
- Optimal Model Averaging: Towards Personalized Collaborative Learning
- Optimal Mixture Weights for Off-Policy Evaluation with Multiple Behavior Policies