Towards Optimal Robustness in Learning-Augmented Paging

2026-05-31

AI Summary

This work addresses the suboptimal robustness of existing learning-augmented paging algorithms, which achieve only a $2H_k + O(1)$ competitive ratio under stochastic assumptions—significantly worse than the optimal $H_k$. By analyzing structural properties of the online optimal algorithm, we introduce “relative prediction budget” as a unifying primitive to guide the design of a new framework that leverages predictions more effectively. Based on this insight, we present the first learning-augmented paging algorithm achieving an optimal robustness bound of $H_k + O(1)$, thereby uncovering the fundamental cause of prior methods’ suboptimality: improper use of predictions. Both theoretical analysis and empirical evaluation demonstrate that our algorithm not only guarantees near-optimal robustness but also substantially outperforms existing approaches in practical scenarios.

Abstract

Learning-augmented paging has been extensively studied in recent years. A key advantage over naive ML-based approaches is \emph{bounded robustness}, which guarantees worst-case performance even when predictions are inaccurate, making these algorithms valuable for real-world systems. Prior work achieves robustness bounds of $2H_k + O(1)$ in the randomized setting, leaving a gap to the optimal competitive ratio $H_k$. In this paper, we study how to close this gap. We begin by reviewing online optimality and proving a new property of the latest $H_k$-competitive algorithm, which facilitates our analysis in the learning-augmented setting. Then, we review existing learning-augmented paging algorithms and introduce a unifying primitive, the \emph{relative prediction budget}, which captures the essence of establishing robustness and reveals that prior algorithms either overuse or underutilize predictions. Guided by the above analysis, we develop a new framework that achieves the best-possible robustness up to an additive constant for learning-augmented paging: $H_k + O(1)$. Experiments further demonstrate strong practical performance.

Problem

• learning-augmented paging
• robustness
• competitive ratio
• online algorithms
• prediction-based algorithms

Innovation

• learning-augmented paging
• robustness
• competitive ratio
• relative prediction budget
• online algorithms