Online Learning with Recency: Algorithms for Sliding-window Streaming Multi-armed Bandits
This work extends streaming bandits to a sliding-window setting, addressing recency effects, which is relevant for online learning with non-stationary data streams.
The paper studies sliding-window streaming multi-armed bandits, where only the most recent W arms are valid. It shows that exact best-arm identification is hard with sublinear memory, but approximate identification is efficient, and provides memory-regret trade-offs for regret minimization.
Motivated by the recency effect in online learning, we study algorithms for single-pass *sliding-window streaming multi-armed bandits (MABs)* in this paper. In this setting, we are given $n$ arms with unknown sub-Gaussian reward distributions and a parameter $W$. The arms arrive in a single-pass stream, and only the most recent $W$ arms are considered valid. The algorithm is required to perform pure exploration and regret minimization with limited memory, defined as the number of stored arms. The model is a natural extension of the streaming multi-armed bandits model (without the sliding window) that has been extensively studied in recent years. We provide a comprehensive analysis of both the pure exploration and regret minimization problems with the model. For pure exploration, we prove that finding the best arm is hard with sublinear memory while finding an approximate best arm admits an efficient algorithm. For regret minimization, we explore a new notion of regret and give sharp memory-regret trade-offs for any single-pass algorithm. We complement our theoretical results with experiments, demonstrating the trade-offs between sample, regret, and memory.