Benchmarks for Deep Off-Policy Evaluation

TL;DR

This paper introduces a comprehensive benchmark for off-policy evaluation in reinforcement learning, providing challenging tasks, datasets, and policies to standardize progress measurement and facilitate comparison of OPE methods.

Contribution

It presents a unified benchmark with diverse high-dimensional control tasks and policies, addressing the lack of standardized evaluation tools in off-policy evaluation research.

Findings

State-of-the-art algorithms evaluated on new benchmark

Benchmark includes challenging high-dimensional control tasks

Open-source datasets and code provided for future research

Abstract

Off-policy evaluation (OPE) holds the promise of being able to leverage large, offline datasets for both evaluating and selecting complex policies for decision making. The ability to learn offline is particularly important in many real-world domains, such as in healthcare, recommender systems, or robotics, where online data collection is an expensive and potentially dangerous process. Being able to accurately evaluate and select high-performing policies without requiring online interaction could yield significant benefits in safety, time, and cost for these applications. While many OPE methods have been proposed in recent years, comparing results between papers is difficult because currently there is a lack of a comprehensive and unified benchmark, and measuring algorithmic progress has been challenging due to the lack of difficult evaluation tasks. In order to address this gap, we present a collection of policies that in conjunction with existing offline datasets can be used for benchmarking off-policy evaluation. Our tasks include a range of challenging high-dimensional continuous control problems, with wide selections of datasets and policies for performing policy selection. The goal of our benchmark is to provide a standardized measure of progress that is motivated from a set of principles designed to challenge and test the limits of existing OPE methods. We perform an evaluation of state-of-the-art algorithms and provide open-source access to our data and code to foster future research in this area.