Cross-Entropy Method for Content Placement and User Association in Cache-Enabled Coordinated Ultra-Dense Networks

Abstract

Due to the high splitting-gain of dense small cells, Ultra-Dense Network (UDN) is regarded as a promising networking technology to achieve high data rate and low latency in 5G mobile communications. In UDNs, each User Equipment (UE) may receive signals from multiple Base Stations (BSs), which impose severe interference in the networks and in turn motivates the possibility of using Coordinated Multi-Point (CoMP) transmissions to further enhance network capacity. In CoMP-based Ultra-Dense Networks, a great challenge is to tradeoff between the gain of network throughput and the worsening backhaul latency. Caching popular files on BSs has been identified as a promising method to reduce the backhaul traffic load. In this paper, we investigated content placement strategies and user association algorithms for the proactive caching ultra dense networks. The problem has been formulated to maximize network throughput of cell edge UEs under the consideration of backhaul load, which is a constrained non-convex combinatorial optimization problem. To decrease the complexity, the problem is decomposed into two suboptimal problems. We first solved the content placement algorithm based on the cross-entropy (CE) method to minimize the backhaul load of the network. Then, a user association algorithm based on the CE method was employed to pursue larger network throughput of cell edge UEs. Simulation were conducted to validate the performance of the proposed cross-entropy based schemes in terms of network throughput and backhaul load. The simulation results show that the proposed cross-entropy based content placement scheme significantly outperform the conventional random and Most Popular Content placement schemes, with with 50% and 20% backhaul load decrease respectively. Furthermore, the proposed cross-entropy based user association scheme can achieve 30% and 23% throughput gain, compared with the conventional N-best, No-CoMP, and Threshold based user association schemes.

Deep Learning Based User Association in Heterogeneous Wireless Networks

“…Under the number of UEs K = 44, the U-Net and the "U- Net(General)" scheme can achieve respectively more than 52% and 35% minimum rate gain. Moreover, both the proposed U-Net scheme and the "U-Net(General)" scheme approach closely the asymptotic optimal GA association scheme in [14], with more than 97% minimum rates under the number of users K = 28. Under the number of users K = 44, the minimum rate of the proposed U-Net scheme and the "U-Net(General)" scheme are respectively 93% and 83%, compared to that of the GA scheme.…”

“…In Fig.3-Fig.5, "U-Net" represents the performance of the loss function L C -based U-Net learning scheme (L C is defined in (13)). "NB" and "GA" represent the performance of the conventional the N-best scheme and the cross-entropy based Genetic Algorithm in [14] respectively. Fig.3 shows the minimum rates of the wireless network achieved by the proposed deep learning schemes, the conventional N-best (NB) association scheme and the asymptotic optimal GA association scheme in [14].…”

“…To generate data to train and test our modified U-Net learning model, the input channel gain matrices have to be labeled. In this paper, a cross-entropy algorithm is applied [14] to obtain labels X l , which is the suboptimal solution of problem (8) by essentially a heuristic search Genetic Algorithm. The training dataset is divided into training set, validation set and testing set.…”

Information Theory for Data Communications and Processing

“…This book, composed of the collection of papers that have appeared in the Special Issue of the Entropy journal dedicated to “Information Theory for Data Communications and Processing”, reflects, in its eleven chapters, novel contributions based on the firm basic grounds of information theory. The book chapters [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 ] address timely theoretical and practical aspects that carry both interesting and relevant theoretical contributions, as well as direct implications for modern current and future communications systems.…”

“…This book focuses on timely and relevant features, and the contributions in the eleven book chapters [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 ], summarized below, address the information theoretical frameworks that have important practical implications.…”