Deep learning-based edge caching for multi-cluster heterogeneous networks
- PDF / 2,208,104 Bytes
- 12 Pages / 595.276 x 790.866 pts Page_size
- 16 Downloads / 237 Views
(0123456789().,-volV)(0123456789().,-volV)
MULTI-ACCESS EDGE COMPUTING ENABLED INTERNET OF THINGS
Deep learning-based edge caching for multi-cluster heterogeneous networks Jiachen Yang1 • Jipeng Zhang1 • Chaofan Ma1 • Huihui Wang2 • Juping Zhang3 • Gan Zheng4 Received: 28 October 2018 / Accepted: 14 January 2019 Springer-Verlag London Ltd., part of Springer Nature 2019
Abstract In this work, we consider a time and space evolution cache refreshing in multi-cluster heterogeneous networks. We consider a two-step content placement probability optimization. At the initial complete cache refreshing optimization, the joint optimization of the activated base station density and the content placement probability is considered. And we transform this optimization problem into a GP problem. At the following partial cache refreshing optimization, we take the time–space evolution into consideration and derive a convex optimization problem subjected to the cache capacity constraint and the backhaul limit constraint. We exploit the redundant information in different content popularity using the deep neural network to avoid the repeated calculation because of the change in content popularity distribution at different time slots. Trained DNN can provide online response to content placement in a multi-cluster HetNet model instantaneously. Numerical results demonstrate the great approximation to the optimum and generalization ability. Keywords DNN HetNets Joint optimization User cluster Content placement
1 Introduction The explosive data growth in online social network has brought us a serious challenge to the design of next-generation network architecture [1, 2]. HetNets and edge caching are two key technologies to meet the ever-growing wireless data demand by increasing the regional spectral efficiency, decreasing the transmission delay and avoiding the use of the limited backhaul capacity [3–7]. The study about the content popularity distribution shows that files requested by the users tend to have a heavy-tailed distribution, i.e., Zipf distribution [8], which means the few very & Jipeng Zhang [email protected] 1
School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China
2
Department of Engineering, Jacksonville University, Jacksonville, FL 32211, USA
3
Nankai University, No. 94 Weijin Road, Nankai District, Tianjin 300071, China
4
Wolfson School of Mechanical, Electrical and Manufacturing Engineering, Loughborough University, Loughborough LE113TU, UK
popular files dominate the requests of the users. Inspired by this fact, cache is introduced to reduce the duplicate file transmission. Beside the optimal cache strategy to improve the system performance, we can also apply the BS sleeping technology to improve the overall network energy efficiency [9, 10]. Studies have shown that BSs are largely underutilized especially at weekends [11]. Inspired by the memory hierarchy in the computer science which introduces the main memory between the CPU and hard disk to reduce the commun
Data Loading...
