Unified Framework of Distributional Regret in Multi-Armed Bandits and Reinforcement Learning
This work provides a principled characterization of regret distribution for bandits and RL, offering a new theoretical tool for analyzing tail risk and instance-dependent behavior.
The paper introduces a unified framework for distributional regret in stochastic multi-armed bandits and episodic reinforcement learning, providing probabilistic guarantees that hold uniformly over all confidence levels. It achieves optimal trade-offs between expected and distributional regret, and for multi-armed bandits, obtains a distributional regret bound of O(√(AT) log(1/δ)), confirming a conjecture by Lattimore & Szepesvári.
We study the distribution of regret in stochastic multi-armed bandits and episodic reinforcement learning through a unified framework. We formalize a distributional regret bound as a probabilistic guarantee that holds uniformly over all confidence levels $δ\in (0,1]$, thereby characterizing the regret distribution across the full range of $δ$. We present a simple UCBVI-style algorithm with exploration bonus $\min\{c_{1,k}/N, c_{2,k}/\sqrt{N}\}$, where $N$ denotes the visit count and $(c_{1,k},c_{2,k})$ are user-specified parameters. For arbitrary parameter sequences, we derive general gap-independent and gap-dependent distributional regret bounds, yielding a principled characterization of how the parameters control the trade-off between expected performance, tail risk, and instance-dependent behavior. In particular, our bounds achieve optimal trade-offs between expected and distributional regret in both minimax and instance-dependent regimes. As a special case, for multi-armed bandits with $A$ arms and horizon $T$, we obtain a distributional regret bound of order $\mathcal{O}(\sqrt{AT}\log(1/δ))$, confirming the conjecture of Lattimore & Szepesvári (2020, Section 17.1) for the first time.