Semigradient-Based Cooperative Caching Algorithm for Mobile Social Networks

Abstract

Wireless caching at users' devices in mobile social network is considered to be a promising solution to alleviate backhaul overload in future wireless networks. However, most of the current works propose caching schemes based on heuristic reasoning and intuition with poor performance or high complexity which are impractical due to individual devices' computing capacity restriction. In this paper, we design a cooperative caching scheme aimed at maximizing hit ratio, incorporating probabilistic modeling of mobility and user interests patterns from mobile social networks. Furthermore, we reformulate this optimization problem into a submodular function maximization and propose a semigradient-based cooperative caching scheme, while this scheme's efficiency is shown to significantly outperform the greedy caching by 99.6%.