Learning the Optimal Network with Context Awareness: Transfer RL Based Network Selection
This chapter focuses on how to realize context-aware network selection. The context information provides a characterization of user-, traffic-, and network-related properties, which is able to enable fine-grained optimization for network selection. Following a similar way with Chap. 4, we can formulate context-aware network selection as an MDP model by generalizing the state to context information. However, the high resolution of context information may lead to large state space, which could result in low learning efficiency. To handle this issue, we employ a transfer learning idea. Specifically, the time–location- dependent periodic changing rule of load statistical distributions is used to realize efficient online network selection via knowledge transfer. Simulation results show that the proposed transfer RL algorithm could achieve better convergence performance by reusing learning experience.
- 2.Pantisano F, Bennis M, Saad W et al (2013) Matching with externalities for context-aware user-cell association in small cell networks. In: IEEE global telecommunications conference (GLOBECOM)Google Scholar
- 8.Basnayaka DA Haas H (2015) Hybrid RF and VLC systems: improving user data rate performance of VLC systems. In: IEEE Vehicular Technology Conference (VTC)Google Scholar
- 10.GigaIR, Infrared Data Association Standards, http://irda.org
- 13.Talvitie E, Singh S (2007) An experts algorithm for transfer learning. In: international joint conference on artificial intelligence (IJCAI)Google Scholar
- 14.Ibrahim M, Khawam K, Tohme S (2010) Congestion games for distributed radio access selection in broadband networks. In: IEEE global telecommunications conference (GLOBECOM)Google Scholar