Beijing University of Posts and Telecommunications

About: Beijing University of Posts and Telecommunications is a education organization based out in Beijing, Beijing, China. It is known for research contribution in the topics: MIMO & Quality of service. The organization has 39576 authors who have published 41525 publications receiving 403759 citations. The organization is also known as: BUPT.

Streaming High-Quality Mobile Video with Multipath TCP in Heterogeneous Wireless Networks

Abstract: The proliferating wireless infrastructures with complementary characteristics prompt the bandwidth aggregation for concurrent video transmission in heterogeneous access networks. Multipath TCP (MPTCP) is an important transport-layer protocol recommended by IETF to integrate different access medium (e.g., Cellular and Wi-Fi). This paper investigates the problem of mobile video delivery using MPTCP in heterogeneous wireless networks with multihomed terminals. To achieve the optimal quality of real-time video streaming, we have to seriously consider the path asymmetry in different access networks and the disadvantages of the data retransmission mechanism in MPTCP. Motivated by addressing these critical issues, this study presents a novel quAlity-Driven MultIpath TCP (ADMIT) scheme that integrates the utility maximization based Forward Error Correction (FEC) coding and rate allocation. We develop an analytical framework to model the MPTCP-based video delivery quality over multiple communication paths. ADMIT is able to effectively integrate the most reliable access networks with FEC coding to minimize the end-to-end video distortion. The performance of ADMIT is evaluated through extensive semi-physical emulations in Exata involving H.264 video streaming. Experimental results show that ADMIT outperforms the reference transport protocols in terms of video PSNR (Peak Signal-to-Noise Ratio), end-to-end delay, and goodput. Thus, we recommend ADMIT for streaming high-quality mobile video in heterogeneous wireless networks with multihomed terminals.

Blockchain Meets Edge Computing: Stackelberg Game and Double Auction Based Task Offloading for Mobile Blockchain

Abstract: Blockchain technology is developing rapidly and has been applied in various aspects, among which there are broad prospects in Internet of Things (IoT). However, IoT mobile devices are restricted in communication and computation due to mobility and portability, so that they can’t afford the high computing cost for blockchain mining process. To solve it, the free resources displayed on non-mining-devices and edge cloud are selected to construct collaborative mining network(CMN) to execute mining tasks for mobile blockchain. Miners can offload their mining tasks to non-mining-devices within a CMN or the edge cloud when there are insufficient resources. Considering competition for resource of non-mining-devices, resource allocation problem in a CMN is formulated as a double auction game, among which Bayes-Nash Equilibrium (BNE) is analyzed to figure out the optimal auction price. When offloading to edge cloud, Stackelberg game is adopted to model interactions between edge cloud operator and different CMNs to obtain the optimal resource price and devices’ resource demands. The mechanism realizes improving the mining utility in mining networks while ensuring the maximum profit of edge cloud operator. Finally, profits of mining networks are compared with an existing mode which only considers offloading to edge cloud. Under the proposed mechanism, mining networks obtain 6.86% more profits on average.

Variational Bayesian Learning for Dirichlet Process Mixture of Inverted Dirichlet Distributions in Non-Gaussian Image Feature Modeling

Abstract: In this paper, we develop a novel variational Bayesian learning method for the Dirichlet process (DP) mixture of the inverted Dirichlet distributions, which has been shown to be very flexible for modeling vectors with positive elements. The recently proposed extended variational inference (EVI) framework is adopted to derive an analytically tractable solution. The convergency of the proposed algorithm is theoretically guaranteed by introducing single lower bound approximation to the original objective function in the EVI framework. In principle, the proposed model can be viewed as an infinite inverted Dirichlet mixture model that allows the automatic determination of the number of mixture components from data. Therefore, the problem of predetermining the optimal number of mixing components has been overcome. Moreover, the problems of overfitting and underfitting are avoided by the Bayesian estimation approach. Compared with several recently proposed DP-related methods and conventional applied methods, the good performance and effectiveness of the proposed method have been demonstrated with both synthesized data and real data evaluations.

Racial Faces in-the-Wild: Reducing Racial Bias by Information Maximization Adaptation Network

Abstract: Racial bias is an important issue in biometric, but has not been thoroughly studied in deep face recognition. In this paper, we first contribute a dedicated dataset called Racial Faces in-the-Wild (RFW) database, on which we firmly validated the racial bias of four commercial APIs and four state-of-the-art (SOTA) algorithms. Then, we further present the solution using deep unsupervised domain adaptation and propose a deep information maximization adaptation network (IMAN) to alleviate this bias by using Caucasian as source domain and other races as target domains. This unsupervised method simultaneously aligns global distribution to decrease race gap at domain-level, and learns the discriminative target representations at cluster level. A novel mutual information loss is proposed to further enhance the discriminative ability of network output without label information. Extensive experiments on RFW, GBU, and IJB-A databases show that IMAN successfully learns features that generalize well across different races and across different databases.

A unified enhanced particle swarm optimization‐based virtual network embedding algorithm

Abstract: SUMMARY Virtual network (VN) embedding is a major challenge in network virtualization. In this paper, we aim to increase the acceptance ratio of VNs and the revenue of infrastructure providers by optimizing VN embedding costs. We first establish two models for VN embedding: an integer linear programming model for a substrate network that does not support path splitting and a mixed integer programming model when path splitting is supported. Then we propose a unified enhanced particle swarm optimization-based VN embedding algorithm, called VNE-UEPSO, to solve these two models irrespective of the support for path splitting. In VNE-UEPSO, the parameters and operations of the particles are well redefined according to the VN embedding context. To reduce the time complexity of the link mapping stage, we use shortest path algorithm for link mapping when path splitting is unsupported and propose greedy k-shortest paths algorithm for the other case. Furthermore, a large to large and small to small preferred node mapping strategy is proposed to achieve better convergence and load balance of the substrate network. The simulation results show that our algorithm significantly outperforms previous approaches in terms of the VN acceptance ratio and long-term average revenue. Copyright © 2012 John Wiley & Sons, Ltd.