Similarity Caching: Theory and Algorithms

This paper focuses on similarity caching systems, in which a user request for an object $o$ that is not the cache can be (partially) satisfied by similar stored notation="LaTeX">$o'$ , at cost of loss utility. Similarity systems effectively employed several application areas, like multimedia retrieval, recommender genome study, and machine learning training/serving. However, despite their relevance, behavior such far from being well understood. In this paper, we provide first comprehensive analysis offline, adversarial, stochastic settings. We show raises significant new challenges, propose dynamic policies with some optimality guarantees. evaluate performance our schemes under both synthetic real traces.