Active Popularity Learning with Cache Hit Ratio Guarantees using a Matrix Completion Committee

Edge caching is a promising technology to facethe stringent latency requirements and back-haul trafficoverloading in 5G wireless networks. However, acquiringthe contents and modeling the optimal cache strategy is achallenging task. In this work, we use an active learningapproach to learn the content popularities since it allowsthe system to leverage the trade-off between explorationand exploitation. Exploration refers to caching new fileswhereas exploitation use known files to cache, to achievea good cache hit ratio. In this paper, we mainly focus tolearn popularities as fast as possible while guaranteeing anoperational cache hit ratio constraint. The effectiveness ofproposed learning and caching policies are demonstratedvia simulation results as a function of variance, cache hitratio and used storage.