基于元学习和二进制粒子群的网络故障特征选择算法

为了降低Wrapper模式网络故障特征选择方法分类算法的计算量,本文提出了一种基于元学习和二进制粒子群(ML-BPSO)的特征选择方法。算法在封装的分类训练中采用元学习方法估算分类精度,并利用BPSO在特征空间中进行全局搜索选出最优特征集。在DARPA数据集上的实验表明本文提出的方法能够显著的降低封装模式特征选择计算量,同时保证了较高的诊断精度和较好的降维效果。     

An energy efficient scheme by exploiting multi-hop D2D communications for 5G networks

The multi-hop Device-to-Device (M-D2D) communication has a potential to serve as a promising technology for upcoming 5G networks. The prominent reason is that the M-D2D communication has the potential to improve coverage, enhanced spectrum efficiency, better link quality, and energy-efficient communication. One of the major challenges for M-D2D communication is the mitigation of interference between the cellular user (CUs) and M-D2D users. Considering this mutual interference constraint, this work investigates the problem of optimal matching of M-D2D links and CUs to form spectrum-sharing partners to maximize overall sum rates of the cell under QoS and energy efficiency (EE) constraints. In this paper, we investigate the interference management for multi-hop (more than one-hop) D2D communication scenarios where we propose a channel assignment scheme along with a power allocation scheme. The proposed channel assignment scheme is based on the Hungarian method in which the channel assignment for M-D2D pairs is done by minimum interference value. The power allocation scheme is based on Binary Particle swarm optimization (BPSO). This scheme calculates the specific power values for all the individual M-D2D links. We have done a comprehensive simulation and the result portrays that our proposed scheme performs better compared to the previous work mentioned in the literature. The results clearly indicate that the proposed scheme enhances the EE of up to 13% by producing the optimal assignment of channels and power for the CUs and M-D2D users.