Showing Your Offline Reinforcement Learning Work: Online Evaluation Budget Matters

TL;DR

This paper emphasizes the importance of online evaluation budgets in comparing deep offline RL algorithms, showing that budget size influences algorithm performance and suggesting reporting methods for more reliable comparisons across domains.

Contribution

It introduces the concept of online evaluation budgets, advocates for reporting performance under varying budgets, and adapts NLP tools for RL evaluation to improve reliability.

Findings

Behavioral Cloning often outperforms offline RL under limited budgets.

Evaluation budget impacts algorithm ranking across domains.

Proposes a new reporting tool for performance estimation.

Abstract

In this work, we argue for the importance of an online evaluation budget for a reliable comparison of deep offline RL algorithms. First, we delineate that the online evaluation budget is problem-dependent, where some problems allow for less but others for more. And second, we demonstrate that the preference between algorithms is budget-dependent across a diverse range of decision-making domains such as Robotics, Finance, and Energy Management. Following the points above, we suggest reporting the performance of deep offline RL algorithms under varying online evaluation budgets. To facilitate this, we propose to use a reporting tool from the NLP field, Expected Validation Performance. This technique makes it possible to reliably estimate expected maximum performance under different budgets while not requiring any additional computation beyond hyperparameter search. By employing this tool, we also show that Behavioral Cloning is often more favorable to offline RL algorithms when working within a limited budget.