A Framework for Fair Evaluation of Variance-Aware Bandit Algorithms

TL;DR

This paper introduces a standardized evaluation framework for multi-armed bandit algorithms, especially variance-aware ones, revealing their advantages in high-uncertainty environments and providing insights into their comparative performance.

Contribution

It presents a reproducible evaluation framework for MAB algorithms and analyzes conditions favoring variance-aware methods over classical algorithms.

Findings

Variance-aware algorithms excel in high-uncertainty settings.

Classical algorithms perform well in separable scenarios.

Extensive tuning benefits classical algorithms more.

Abstract

Multi-armed bandit (MAB) problems serve as a fundamental building block for more complex reinforcement learning algorithms. However, evaluating and comparing MAB algorithms remains challenging due to the lack of standardized conditions and replicability. This is particularly problematic for variance-aware extensions of classical methods like UCB, whose performance can heavily depend on the underlying environment. In this study, we address how performance differences between bandit algorithms can be reliably observed, and under what conditions variance-aware algorithms outperform classical ones. We present a reproducible evaluation designed to systematically compare eight classical and variance-aware MAB algorithms. The evaluation framework, implemented in our Bandit Playground codebase, features clearly defined experimental setups, multiple performance metrics (reward, regret, reward distribution, value-at-risk, and action optimality), and an interactive evaluation interface that supports consistent and transparent analysis. We show that variance-aware algorithms can offer advantages in settings with high uncertainty where the difficulty arises from subtle differences between arm rewards. In contrast, classical algorithms often perform equally well or better in more separable scenarios or if fine-tuned extensively. Our contributions are twofold: (1) a framework for systematic evaluation of MAB algorithms, and (2) insights into the conditions under which variance-aware approaches outperform their classical counterparts.