Machine Learning for Wireless Caching
Supervisor: Dr Maged Elkashlan
Research group(s): NetworksOne of the promising techniques to solve backhaul bottleneck issue is a paradigm of proactive caching at network edge that exploits the recent advances in storage, context-aware networking, big data, and heterogeneous networks. The nodes at the edge of the network (closer to the user) can predict the popular content, store the content, and deliver it when requested by a user. However, most of the work in literature does not consider the dynamic nature of the popular content and assumes that content popularity is fixed. Since the content popularity is time-varying, caching can only be optimized if the fresh view of the system is maintained. This requires huge data collection, data processing, and statistical inference from this data. This can only be realized by integrating machine learning (ML) tools across the wireless infrastructure and end-user devices. This project will design new proactive cooperative caching techniques that combine ML for a time-varying content popularity. We will use ML algorithms such as reinforcement learning, neural networks, and deep learning to predict popular content and update them with time. Therefore, the objective of this research is to design and develop cooperative content caching and delivery policies using ML techniques for proactive caching at network edge.