Dataset Information

Variance Regularized Counterfactual Risk Minimizationvia Variational Divergence Minimization.

ABSTRACT: Off-policy learning, the task of evaluating and improving policies using historic data collected from a logging policy, is important because on-policy evaluation is usually expensive and has adverse impacts. One of the major challenge of off-policy learning is to derive counterfactual estimators that also has low variance and thus low generalization error. In this work, inspired by learning bounds for importance sampling problems, we present a new counterfactual learning principle for off-policy learning with bandit feedbacks. Our method regularizes the generalization error by minimizing the distribution divergence between the logging policy and the new policy, and removes the need for iterating through all training samples to compute sample variance regularization in prior work. With neural network policies, our end-to-end training algorithms using variational divergence minimization showed significant improvement over conventional baseline algorithms and is also consistent with our theoretical results.

SUBMITTER: Wu H

PROVIDER: S-EPMC7419136 | biostudies-literature | 2018

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Variance Regularized Counterfactual Risk Minimizationvia Variational Divergence Minimization.

Wu Hang H Wang May D MD

Proceedings of machine learning research 20180101

Off-policy learning, the task of evaluating and improving policies using historic data collected from a logging policy, is important because on-policy evaluation is usually expensive and has adverse impacts. One of the major challenge of off-policy learning is to derive counterfactual estimators that also has low variance and thus low generalization error. In this work, inspired by learning bounds for importance sampling problems, we present a new counterfactual learning principle for off-policy ...[more]

PMID: 32789292

Similar Datasets

Project description:BackgroundA Guide for Healthcare Professionals (HCP Guide) and patient alert card (PAC) for atezolizumab as additional risk minimization measures for physicians were distributed to raise awareness and help in the detection and management of immune-related adverse drug reactions.ObjectivesThe main objective of this study was to assess the receipt, knowledge, and behaviors of physicians regarding the atezolizumab HCP Guide and PAC.MethodsA multi-country, one-wave, observational, cross-sectional, web-based, self-reported physician survey was conducted to assess the level of knowledge of key messages related to immune-related adverse drug reactions summarized in the atezolizumab HCP Guide and PAC among physicians (oncologists, pulmonologists, and urologists) prescribing atezolizumab in six European countries (Denmark, Germany, Italy, Spain, Sweden, and the UK). Responses regarding the receipt, understanding and use of the materials, and knowledge and behavior related to the HCP Guide and PAC are presented as percentages and continuous scores scaled out of 100 points, with corresponding 95% confidence intervals (CIs).ResultsAmong 313 physicians (255 oncologists, 30 pulmonologists, and 28 urologists), 77.4% received the HCP Guide and 74.2% the PAC. The HCP Guide was read by 71.3% of the 267 physicians who received the materials, and the mean usage score was 69.5 (95% CI 66.0-72.9), and 57.1% of physicians had scores ≥ 70. The HCP Guide was completely understood by 85.4% of physicians who had read it. Mean knowledge scores were 63.9 (95% CI 62.1-65.7) and 39.4% of physicians had correct knowledge scores ≥ 70. Mean knowledge scores were 66.8 (95% CI 64.9-68.7) for receipt of both the HCP Guide and PAC, 59.4 (95% CI 55.5-63.4) for one of the materials, and 60.8 (95% CI 55.4-66.2) for having received none of the materials. Mean behavior scores were 78.9 (95% CI 76.8-81.0), and 74.8% of physicians had behavior scores ≥ 70. The mean behavior score was 79.0 (95% CI 76.5-81.5) for those who received both the HCP Guide and PAC, 76.9 (95% CI 72.2-81.5) for receipt of one of the materials, and 81.5 (95% CI 75.0-88.0) for those who received none of the materials.ConclusionsThe study assessed the effectiveness of the atezolizumab additional risk minimization educational materials among physicians in six European countries, using process indicators. The educational materials reached over 70% of target physicians, 57.1% of whom reported using them. Knowledge and behavior related to immune-related adverse drug reactions for atezolizumab were no better in those who received the additional risk minimization educational materials. The results support the safe use of atezolizumab by these physician groups and contributed to the European Medicines Agency permitting removal of the HCP Guide.

Dataset Information

Variance Regularized Counterfactual Risk Minimizationvia Variational Divergence Minimization.

Publications

Variance Regularized Counterfactual Risk Minimizationvia Variational Divergence Minimization.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets