Cross-Validated Off-Policy Evaluation
In this paper, we study the problem of estimator selection and hyper-parameter tuning in off-
policy evaluation. Although cross-validation is the most popular method for model selection …
policy evaluation. Although cross-validation is the most popular method for model selection …
Pessimistic model selection for offline deep reinforcement learning
CHH Yang, Z Qi, Y Cui… - Uncertainty in Artificial …, 2023 - proceedings.mlr.press
Abstract Deep Reinforcement Learning (DRL) has demonstrated great potentials in solving
sequential decision making problems in many applications. Despite its promising …
sequential decision making problems in many applications. Despite its promising …
[PDF][PDF] A Rigorous Risk-aware Linear Approach to Extended Markov Ratio Decision Processes with Embedded Learning.
A Zadorojniy, T Osogami, O Davidovich - IJCAI, 2023 - ijcai.org
We consider the problem of risk-aware Markov Decision Processes (MDPs) for Safe AI. We
introduce a theoretical framework, Extended Markov Ratio Decision Processes (EMRDP) …
introduce a theoretical framework, Extended Markov Ratio Decision Processes (EMRDP) …
AutoOPE: Automated Off-Policy Estimator Selection
N Felicioni, M Benigni, MF Dacrema - arXiv preprint arXiv:2406.18022, 2024 - arxiv.org
The Off-Policy Evaluation (OPE) problem consists of evaluating the performance of
counterfactual policies with data collected by another one. This problem is of utmost …
counterfactual policies with data collected by another one. This problem is of utmost …
[PDF][PDF] Large-scale open dataset, pipeline, and benchmark for bandit algorithms
Y Saito, S Aihara, M Matsutani… - arXiv preprint arXiv …, 2020 - dynamicdecisions.github.io
We build and publicize the Open Bandit Dataset to facilitate scalable and reproducible
research on bandit algorithms. It is especially suitable for off-policy evaluation (OPE), which …
research on bandit algorithms. It is especially suitable for off-policy evaluation (OPE), which …
A Fine-grained Analysis of Fitted Q-evaluation: Beyond Parametric Models
J Wang, Z Qi, RKW Wong - arXiv preprint arXiv:2406.10438, 2024 - arxiv.org
In this paper, we delve into the statistical analysis of the fitted Q-evaluation (FQE) method,
which focuses on estimating the value of a target policy using offline data generated by …
which focuses on estimating the value of a target policy using offline data generated by …
Is Separately Modeling Subpopulations Beneficial for Sequential Decision-Making?
I Lee - Operations Research, 2023 - pubsonline.informs.org
In recent applications of Markov decision processes (MDPs), it is common to estimate
transition probabilities and rewards from transition data. In healthcare and some other …
transition probabilities and rewards from transition data. In healthcare and some other …
Optimizing Warfarin Dosing Using Contextual Bandit: An Offline Policy Learning and Evaluation Method
Y Huang, CA Downs, AM Rahmani - arXiv preprint arXiv:2402.11123, 2024 - arxiv.org
Warfarin, an anticoagulant medication, is formulated to prevent and address conditions
associated with abnormal blood clotting, making it one of the most prescribed drugs globally …
associated with abnormal blood clotting, making it one of the most prescribed drugs globally …
Leveraging Factored Action Spaces for Off-Policy Evaluation
Off-policy evaluation (OPE) aims to estimate the benefit of following a counterfactual
sequence of actions, given data collected from executed sequences. However, existing OPE …
sequence of actions, given data collected from executed sequences. However, existing OPE …
Off environment evaluation using convex risk minimization
Applying reinforcement learning (RL) methods on robots typically involves training a policy
in simulation and deploying it on a robot in the real world. Because of the model mismatch …
in simulation and deploying it on a robot in the real world. Because of the model mismatch …