What are the Statistical Limits of Offline RL with Linear Function Approximation?

Authors: Ruosong Wang, Dean Foster, Sham M. Kakade

ICLR 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical Perhaps surprisingly, our main result shows that even if: i) we have realizability in that the true value function of every policy is linear in a given set of features and 2) our off-policy data has good coverage over all features (under a strong spectral condition), any algorithm still (information-theoretically) requires a number of offline samples that is exponential in the problem horizon to nontrivially estimate the value of any given policy.
Researcher Affiliation Collaboration Ruosong Wang Carnegie Mellon University ruosongw@andrew.cmu.edu Dean P. Foster University of Pennsylvania and Amazon dean@foster.net Sham M. Kakade University of Washington, Seattle and Microsoft Research sham@cs.washington.edu
Pseudocode Yes Algorithm 1 Least-Squares Policy Evaluation
Open Source Code No The paper does not provide any statement or link indicating the release of open-source code for the methodology described.
Open Datasets No The paper constructs theoretical
Dataset Splits No The paper is theoretical and does not involve empirical validation on datasets requiring explicit train/validation/test splits.
Hardware Specification No The paper is theoretical and does not describe any hardware used for experiments.
Software Dependencies No The paper is theoretical and does not describe specific software dependencies with version numbers.
Experiment Setup No The paper is theoretical and focuses on statistical limits and algorithm analysis, not on empirical experiment setup with hyperparameters.