Importance Sampling Via the Estimated Sampler

By Masayuki Henmi, Ryo Yoshida, Shinto Eguchi. 2007

40 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.

Black-box Off-policy Estimation for Infinite-Horizon Reinforcement Learning (2020, background): “Interestingly, the authors show that it is beneﬁcial to estimate propensity scores from data without using knowledge of the behavior distribution (called proposal distribution in that paper), even if it is available; see also [13] for related arguments.”
Learning to Draw Samples: With Application to Amortized MLE for Generative Adversarial Learning (2016, neutral): “…there exist methods that use no explicit information of q ( x ) , which, seemingly counter-intuitively, give better asymptotic variance or converge rates than the typical IS that uses the proposal information (e.g., Liu & Lee, 2016; Briol et al., 2015; Henmi et al., 2007; Delyon & Portier, 2014).”
Importance Sampling Policy Evaluation with an Estimated Behavior Policy (2018, methodology): “Methods related to RIS have been studied for Monte Carlo integration (Henmi et al., 2007; Delyon & Portier, 2016) and causal inference (Hirano et al.”
Off-Policy Evaluation and Learning from Logged Bandit Feedback: Error Reduction via Surrogate Policy (2018, methodology): “This idea dates back to the “estimated sampler” technique in statistics [4, 7].”
Black-box Importance Sampling (2016, methodology): “Henmi et al. (2007); Delyon and Portier (2014) showed that certain types of approximation ˆ q can improve, rather than deteriorate, the performance compared with the weight with the exact q .”