On the Exploration and Exploitation Trade-off in Cooperative Caching-enabled Networks

Edge-caching is considered as a promising solution to address network congestion and reduce delivery latency in the future by bringing the relevant contents close to users. In this context, the commonly used notion involves the storage of the most popular contents in the cache, while consequently increasing the cache hit ratio (CHR). In the majority of prior works, the content popularity is assumed to be perfectly known and often a priori. However, in reality, the content popularity has to be explored especially for uncertain contents, such as new entrants and fast varying items. In this paper, we develop a framework to analyze the joint exploration and exploitation trade-off by caching both popular and uncertain contents to enable more efficient content caching. Particularly, we formulate an optimization problem to maximize the trade-off between exploration and exploitation subject to the maximum storage capacity, guaranteed CHR, and back-haul energy budget constraints. Further, we solve the formulated mixed-integer combinatorial problem using branch-and-bound optimizer by relaxing the binary to box constraints. The superiority in performance of the proposed method over the state-of-the-art is demonstrated in terms of the CHR and back-haul energy on a realistic Movie-lens dataset.