Collaborative Caching in Edge Computing via Federated Learning and Deep Reinforcement Learning

By deploying resources in the vicinity of users, edge caching can substantially reduce latency for users to retrieve content and relieve pressure on backbone network. Due capacity limitation dynamic nature user requests, how allocate reasonably must be considered. Some studies improve network performance by predicting popularity actively most popular content, thereby ignoring privacy security issues caused need collect information at central unit. To this end, a collaborative strategy based federated learning is proposed. First, used make distributed predictions preferences nodes develop an effective policy. Then, problem allocating optimize cost video providers formulated as Markov decision process, reinforcement method decisions. Compared with several basic strategies terms cache hit rate, transmission delay, cost, simulation results show that proposed reduces providers, has higher rate lower average delay.