Mobile users can open program in new tab for better viewing.

Open program in new tab

Day 1 24/08/2019
Room #1

Opening Ceremony (Moderator: Hai Deng) 07:00 - 07:10

Welcome and Opening Remarks by President: Bing Chen

Keynote I (Keynote Chair: Qihui Wu) 07:10 - 07:55

“Collaborative Task Execution in Advanced Edge Computing Environments”--Prof. Jiannong Cao, Hong Kong Polytechnic University, China.

Coffee Break 07:55 - 08:30

(Out of Meeting Room)

Keynote II (Keynote Chair: Kun Zhu) 08:30 - 09:15

“Accessing From the Sky: UAV Communications for 5G and Beyond” --by Prof. Rui Zhang, National University of Singapore, Singapore

Keynote III (Keynote Chair: Kun Zhu) 09:15 - 10:00

“Network Centrality as Statistical Inference in Large Networks” – by Prof. Chee Wei Tan, City University of Hong Kong, China
Room #2

Session 1.1 (Session Chair: Feng Hu) 12:00 - 13:40

Wireless Network
12:00 - 12:15
High-Dimensional Data Anomaly Detection Framework Based on Feature Extraction of Elastic Network

Although appropriate feature extraction can improve the performance of anomaly detection, it is a challenging task due to the complex interaction between features, the mixture of irrelevant features and relevant features, and the unavailability of data tags. When conventional anomaly detection methods deal with the problem of high dimensional data, the performance of anomaly detection will be degraded due to the existence of irrelevant features. This paper proposed a method of feature extraction and anomaly detection for high dimensional data based on elastic network, which can filter irrelevant features and improve the accuracy and efficiency of anomaly detection. In this paper, an outlier scoring method was used to score the outliers of the original data, and then outliers and the original data were input into the elastic network for sparse regression. Those irrelevant features to abnormal data are ignored after extraction. Finally, high-dimensional data are detected efficiently according to extracted features. In the experimental stage, we used the high-dimensional anomaly dataset provided by ODDS to detect the performance of the proposed method based on detection accuracy (AUC), ROC curve, feature number, convergence speed and other indicators. The results show that the proposed method not only can effectively extract the features related to high-dimensional anomaly data, but also the detection accuracy of outliers has been greatly improved.
Authors: yang shen (State Grid Liaoning Electric Power Supply Co, LTD), jue bo (State Grid Liaoning Electric Power Supply Co, LTD), kexin li (College of Computer Science and Technology, Nanjing University of Aeronautics and Astro-nautics), shuo chen (State Grid Liaoning Electric Power Supply Co, LTD), lin qiao (State Grid Liaoning Electric Power Supply Co, LTD), jing li (College of Computer Science and Technology, Nanjing University of Aeronautics and Astro-nautics),
Hide Authors & Abstract

Show Authors & Abstract
12:15 - 12:30
A Drone Formation Transformation Approach

In the process of performing fixed-wing drone formations, it is usually necessary to perform a variety of formations according to mission requirements or environmental changes. However, performing such formation transformation during formation flight will face many technical challenges. In this paper, we first present a Six-Tuple State Coherence (STSC) model for fixed-wing drone formations, and based on this model, the definition of drone formation transformation is given. Moreover, a drone formation change algorithm (DFCA) is proposed. When a new formation is needed, the master node first adopts the centralized Hungarian algorithm to determine the location allocation scheme of the new formation, and then each node calculates and executes dubins paths distributedly to maintain the consistency of the formation yaw angle, and finally adjusts the speed of the nodes to ensure the formation of STSC. The prototype system conforming to DFCA algorithm is implemented on OMNET++ platform, and numerous simulation experiments are carried out. The experimental results show the feasibility of the DFCA algorithm and show that it can control the drone formation transformation at a lower cost.
Authors: chenghao jin (Nanjing University of Aeronautics and Astronautics), bing chen (Nanjing University of Aeronautics and Astronautics), feng hu (Nanjing University of Aeronautics and Astronautics),
Hide Authors & Abstract

Show Authors & Abstract
12:30 - 12:45
The principle and design of Separate fingerprint identification system

In this paper, a separate fingerprint identification system is proposed. In the beginning, system introduction is described. Then the working mode and the communication protocols are conducted. Separate fingerprint identification system has the modules of fingerprint entry, image processing, fingerprint contrasting, fingerprint searching and module storing without the management of Upper Computer. It can make up a separate fingerprint identification system or being a integrated outer equipment with the help of corresponding fingerprint sensor. Finally, module direction system is discussed.
Authors: Liu Mengmeng,
Hide Authors & Abstract

Show Authors & Abstract
12:45 - 13:00
LTE antenna port number detection algorithm based on channel estimation and piecewise linear regression

In LTE system, blind detection of traditional antenna port number detec-tion generated a lot of computational redundancy and delay. To solve this problem, an improved detection algorithm based on channel estimation and piecewise linear regression is proposed. This algorithm fits the phase in-formation of channel state and determines the number of antenna ports. The problem of decision error caused by phase jump is solved by piece-wise. The theoretical analysis and simulation results show that the pro-posed algorithm has the advantages of low complexity and delay.
Authors: Jiang pengchun (13368124190), Zhou Mu (13983850201),
Hide Authors & Abstract

Show Authors & Abstract
13:00 - 13:15
Topology Sensing in Wireless Networks by Leveraging Symmetrical Connectivity

With the popularization of wireless networks, the role of machine intelligence is becoming more and more important, where the core is that the network needs to make its own decisions through learning. Topology sensing is a fundamental issue in the field of network intellectualization, but most of the related existing studies have focused on wired networks, while the characteristics of wireless networks are relatively few investigated. In this paper, a wireless channel-oriented topology sensing method based on Hawkes process modeling is proposed for the wireless network with symmetrical connectivity. Simulation are carried out to demonstrate that how to combine wireless channel with Hawkes process and how to further process the results to improve performance.
Authors: Zitong Liu (Key Laboratory of Dynamic Cognitive System of Electromagnetic Spectrum Space, Ministry of Industry and Information Technology, College of Electronics and Information Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China), Jiachen Sun (College of Communications Engineering, Army Engineering University, Nanjing 210007, China), Feng Shen (Key Laboratory of Dynamic Cognitive System of Electromagnetic Spectrum Space, Ministry of Industry and Information Technology, College of Electronics and Information Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China), Guoru Ding (College of Communications Engineering, Army Engineering University, Nanjing 210007, China), Qihui Wu (Key Laboratory of Dynamic Cognitive System of Electromagnetic Spectrum Space, Ministry of Industry and Information Technology, College of Electronics and Information Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China),
Hide Authors & Abstract

Show Authors & Abstract
13:15 - 13:30
Multi-destination two-hop untrusted AF relay networks with destination-aided cooperative jamming

We consider a multi-destination two-hop untrusted amplifyand-forward (AF) relay networks, where each node is equipped with a single antenna and the confidential information communication needs the aid of the untrusted relay over the Nakagami-m channel. Because the relay is energy-constrained, the relay needs to harvest energy from the received information and the jamming signals by applying the power splitting relaying (PSR) protocol. The confidential information can be protected from the untrusted relay eavesdropping with destination-aided cooperative jamming. We focus on the secure and reliable performance of the presented system. The secrecy outage probability (SOP) and the connection outage probability (COP) are specially examined, which mainly show in the closed-form expressions of SOP and COP. In addition, the effective secrecy throughput (EST) performance is also investigated to comprehensively measure the secure and reliable performance. Moreover, we also present the asymptotic analysis of EST at the high signal-to-noise ratio (SNR). The Monte Carlo simulation is applied to validate the accuracy of the derived expressions and reveals the effects of different parameters, such as the transmit SNR, the power allocation factor, the fading factor and other parameters on the EST.
Authors: Hui Shi (Army Engineering University of PLA), Wei Yang (Army Engineering University of PLA), Yue Cai (Army Engineering University of PLA), Yong Jia (Army Engineering University of PLA), Wen Yang (Army Engineering University of PLA),
Hide Authors & Abstract

Show Authors & Abstract
13:30 - 13:45
Secrecy Sum Rate for Two-way Untrusted Relay in SCMA Networks

Sparse code multiple access (SCMA) is a novel non-orthogonal multiple access technology that combines the concepts of CDMA and OFDMA. The advantages of SCMA include high capacity, low time delay and high transfer rate.The information security is also very important in 5G network. Relay is essential to be used for long distance cooperative transmission. In this paper, we consider a two-way relay system that each user pair can only communicate through an untrusted intermediate relay. We regard the intermediate relay as an eavesdropper and the confidential information must be kept secret to it. The security performance of the system is analyzed in this paper and the reasonable relay amplify-andforward scheme is analyzed theoretically. In order to maximize the sum security capacity, a subcarrier assignment algorithm based on matching theory is proposed in this paper. Finally, the theoretical analysis is verified by simulation. Simulation results show that security performance is improved significantly.
Authors: Yiteng Huang (Communication Research Center, Harbin Institute of Technology, China), Shuai Han (Communication Research Center, Harbin Institute of Technology, China), Shizeng Guo (Communication Research Center, Harbin Institute of Technology, China), Ming Li (China Academy of space technology, Beijing, China, 100094), Zhiqiang Li (Communication Research Center, Harbin Institute of Technology, China),
Hide Authors & Abstract

Show Authors & Abstract
13:45 - 14:00
Improving complex network controllability via link prediction

Complex network is a network structure composed of a large number of nodes and complex relationships between these nodes. Using complex network can model many systems in real life. The individual in the system corresponds to the node in the network and the relation-ships between these individuals correspond to the edge in the network. The controllability of complex networks is to study how to enable the network to arrive at the desired state from any initial state by external input signals. The external input signals transmit to the whole network through some nodes in the network, and these nodes are called driver node. For the study of controllability of complex network, it is mainly to judge whether the net-work is controllable or not and how to select the appropriate driver nodes. We call the diffi-culty of controlling network controllability. If a network has a high controllability, the net-work will be easy to control However, complex networks are vulnerable to attacks that can cause declining of controllability, Therefore, we propose in this paper a link prediction-based method to make the network more robust to different modes of attacking. Through experiments we have validated the effectiveness of the proposed method.
Authors: Ran Wei (Nanjing University of Aeronautics and Astronautics), Weiwei Yuan (Nanjing University of Aeronautics and Astronautics), Donghai Guan (Nanjing University of Aeronautics and Astronautics), Asad Masood Khattak (Zayed University), Muhammad Fahim (Innopolis University),
Hide Authors & Abstract

Show Authors & Abstract

Coffee Break 13:40 - 14:10

(Out of Meeting Room)

Session 2.1 (Session Chair: Yongliang Sun) 14:10 - 15:50

Big Data and Internet of Things
14:15 - 14:30
Predicting Socio-economic Levels of Individuals via App Usage Records

The socio-economic level of an individual is an indicator of the education, purchasing power and housing. Accurate and proper prediction of the individuals is of great significance for market campaign. However, the previous approaches estimating the socio-economic status of an individual mainly rely on census data which demands a great quantity of money and manpower. In this paper, we analyse two datasets: App usage records and occupation data of individuals in a metropolis of China. We divide the individuals into 4 socio-economic levels according to their occupations. Then, we propose a low-cost socio-economic level classification model constructed with machine learning algorithm. Our predictive model achieves a high accuracy over 80%. Our results show that the features extracted from user’s App usage records are valuable indicators to predict the socio-economics levels of individuals.
Authors: Yi Ren (Sun Yat-sen University), Weimin Mai (Sun Yat-sen University), Yong Li (Tsinghua University), Xiang Chen (Sun Yat-sen University),
Hide Authors & Abstract

Show Authors & Abstract
14:30 - 14:45
Multiple Task Assignment for Cooperating Homogeneous Unmanned Aerial Vehicles

Using multiple unmanned aerial vehicle (UAV) to perform some tasks cooperatively has received growing attention in recent years. Task assignment is one of difficult problem in mission planning. This paper addresses the problem of assigning multiple tasks to cooperating homogeneous UAVs, while minimizing the total cost. Multiple task assignment problem for cooperating homogeneous UAVs is considered a combinatorial optimization problem. In this paper, we propose a centralized task assignment scheme based on minimum spanning tree (MST). The MST-based scheme involves three phases. In the first phase, we use the Kruskal’s algorithm to generate a minimum spanning tree for UAVs and tasks. In the second phase, we first calculate the sum of the weights of the minimum spanning tree which is taken as the lower bound of the total cost. Then, based on the result of lower bound, we use the breadth first search (BFS) algorithm to assign tasks to each UAV and get the initial solution of the task assignment. Finally, the third phase involves the Pareto-optimization improvement in the generated solution in the second phase. For a single UAV, we use the Christofides’ algorithm to calculate the total cost of completing the assigned task. The simulation results show that the scheme can solve the homogeneous multi-UAV cooperative task assignment problem effectively
Authors: Li Li (Nanjing University of Aeronautics and Astronautics), Xiangping Bryce Zhai (Nanjing University of Aeronautics and Astronautics), Bing Chen (Nanjing University of Aeronautics and Astronautics), Congduan Li (Sun Yat-Sen University),
Hide Authors & Abstract

Show Authors & Abstract
14:45 - 15:00
Design of Overall Framework of Self-Service Big Data Governance for Power Grid

At present, power grid companies have not formed a complete data quality control system and a comprehensive and effective data quality assurance mechanism, which restricts the deep mining of data value. In this paper, based on the full-service unified data center of power Grid Company, a general framework of self-service power grid big data governance is presented. Firstly, the related work of big data governance is reviewed; secondly, the architecture and characteristics of the grid company's full-service unified data center are analyzed; thirdly, the related requirements of self-service grid big data governance are analyzed; fourthly, the overall framework of self-service grid big data governance is proposed. Compared with other general big data governance framework, the proposed framework considers the specification of power grid as well as the features of self-service, making the framework more feasible.
Authors: Lin Qiao (State Grid Liaoning Electric Power Co., Ltd. Shenyang 110000, P.R.China), Qiaoni Zhou (State Grid Liaoning Electric Power Co., Ltd. Shenyang 110000, P.R.China), Chunhe Song (Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, 110016, China), Hao Wu (State Grid Liaoning Electric Power Co., Ltd. Shenyang 110000, P.R.China), Biqi Liu (State Grid Liaoning Electric Power Co., Ltd. Shenyang 110000, P.R.China), Shimao Yu (Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, 110016, China),
Hide Authors & Abstract

Show Authors & Abstract
15:00 - 15:15
Data cleaning Based on Multi-sensor Spatiotemporal Correlation

Sensor-based condition monitoring systems are becoming an important part of modern industry. However, the data collected from sensor nodes are usually unreliable and inaccurate. It is very critical to clean the sensor data before using them to detect actual events occurred in the physical world. Popular data cleaning methods, such as moving average and stacked denoise autoencoder, cannot meet the requirements of accuracy, energy efficiency or computation limitation in many sensor related applications. In this paper, we propose a data cleaning method based on multi-sensor spatiotemporal correlation. Specifically, we find out and repair the abnormal data according to the correlation of sensor data in adjacent time and adjacent space. Real data based simulation shows the effectiveness of our proposed method.
Authors: Baozhu Shao (Liaoning Electric Power Research Institute, State Grid Liaoning Electric Power Co., Ltd. Shenyang 110000, P.R.China), Chunhe Song (Shenyang Institute of automation, Chinese Academy of Sciences, Shenyang 110016, P.R.China), Zhongfeng Wang (Shenyang Institute of automation, Chinese Academy of Sciences, Shenyang 110016, P.R.China), Zhexi Li (Shenyang power supply company, State Grid Liaoning Electric Power Co., Ltd. Shenyang 110000, P.R.China), Shimao Yu (Shenyang Institute of automation, Chinese Academy of Sciences, Shenyang 110016, P.R.China), Peng Zeng (Shenyang Institute of automation, Chinese Academy of Sciences, Shenyang 110016, P.R.China),
Hide Authors & Abstract

Show Authors & Abstract
15:15 - 15:30
Distributed Hierarchical Fault Diagnosis Based on Sparse Auto-encoder and Random Forest

For the diagnosis of large-scale local devices, the traditional centralized fault diagnosis systems are becoming incompetent to meet the requirement of real-time monitoring. This paper proposes the Distributed hierarchical Fault Diagnosis System (DFDS). Specifically, DFDS implements fault monitoring by an improved Sparse Auto-Encoder (SAE) on the monitor layer, classifies faults and identifies unknown faults by an improved random forest on the classification layer, learns new knowledge and updates the system on the decision layer. We apply DFDS in the laboratory data of Case Western Reserve University to verify the efficiency of the proposed system. The experimental results show that our method can accurately detect the fault and accurately identify the fault type.
Authors: Tong Li (Liaoning Electric Power Research Institute, State Grid Liaoning Electric Power Co., Ltd. Shenyang 110000, P.R.China), Chunhe Song (Shenyang Institute of automation, Chinese Academy of Sciences, Shenyang 110016, P.R.China), Yang Liu (Liaoning Electric Power Research Institute, State Grid Liaoning Electric Power Co., Ltd. Shenyang 110000, P.R.China), Zhongfeng Wang (Shenyang Institute of automation, Chinese Academy of Sciences, Shenyang 110016, P.R.China), Shimao Yu (Shenyang Institute of automation, Chinese Academy of Sciences, Shenyang 110016, P.R.China), Shanting Su (Nanjing University of Aeronautics and Astronautics, Nanjing, China),
Hide Authors & Abstract

Show Authors & Abstract
15:30 - 15:45
A Data Quality Improvement Method Based on the Greedy Algorithm

High-quality data is the premise of efficient data analysis and mining. Data quality can be indicated by a set of indicators, and some methods have been proposed for data quality improvement by improving one or more data quality indicators. However, there is few work to discuss the impact of the processing order of data quality indicators on the overall data quality. In this paper, first, some data quality indicators and their improvement methods are given, second, the impact of the processing order of data quality indicators on the overall data quality is discussed, and then a novel data quality im-provement method based on the greedy algorithm is proposed. Experiment results have been shown that the proposed method can improves the data quality while reducing the time and computing costs.
Authors: Zhongfeng Wang (Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, 110016, China), Yatong Fu (Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, 110016, China), Chunhe Song (Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, 110016, China), Weichun Ge (State Grid Liaoning Electric Power Co., Ltd. Shenyang 110000, P.R.China), Lin Qiao (State Grid Liaoning Electric Power Co., Ltd. Shenyang 110000, P.R.China), Hongyu Zhang (State Grid Liaoning Electric Power Co., Ltd. Shenyang 110000, P.R.China),
Hide Authors & Abstract

Show Authors & Abstract
15:45 - 16:00
Research of Lightweight Encryption Algorithm Based on AES and Chaotic Sequences for Narrow-Band Internet of Things

a lightweight encryption algorithm based on AES and chaotic sequences for Narrow-Band Internet of Things(NB-IoT) is proposed to solve the data security which maybe appear in NB-IoT. Firstly, we reduce the number of rounds of AES and combine the three steps of ‘SubBytes’, ‘ShiftRows’, and ‘MixColumns’ to improve time efficiency. Secondly, in order to make the algorithm more sucure, we use Logistic and Tent chaotic systems to generate dynamic keys for encryption. Finally, the keys generated by the chaotic systems can encrypt the remaining bytes of the plaintext so that the length of the plaintext is equal to ciphertext. We run LCHAOSAES_128 on the NB-IoT encryption model in application environment proposed by us, and evaluate that LCHAOSAES_128 is suitable for NB-IoT through the experimental results and theoretical analysis.
Authors: Lianmin Shi (Suzhou Institute of Trade & Commerce), Yihuai Wang (Soochow University), Rongyuan Jia (Soochow University), Tao Peng (Soochow University), Jianwu Jiang (Soochow University), Shilang Zhu (Soochow University),
Hide Authors & Abstract

Show Authors & Abstract
16:00 - 16:15
Energy Efficiency Maximization for Green Cognitive Internet of Things with Energy Harvesting

In this paper, a green cognitive Internet of Things (CIoT) has been proposed to collect the radio frequency (RF) energy of primary user (PU) by using energy harvesting. The CIoT nodes are divided into two independent groups to perform spectrum sensing and energy harvesting simultaneously in the sensing slot. The energy efficiency of the CIoT is maximized by through jointly optimizing sensing time, number of sensing nodes and transmission power. The suboptimal solution to the optimization problem is achieved using a joint optimization algorithm based on alternating direction optimization. Simulation results have indicated that the optimal solution is existed and the green CIoT outperforms the traditional scheme.
Authors: xin liu (Dalian University of Technology), Xueyan Zhang (Dalian University of Technology), Weidang Lu (Zhejiang University of Technology), Mudi Xiong (Dalian Maritime University),
Hide Authors & Abstract

Show Authors & Abstract
Room #3

Session 1.2 (Session Chair: Xin Zhang) 12:00 - 13:40

Communication
12:00 - 12:15
Rectangular waveguide design optimization by sequential nonlinear programming and genetic algorithm

Rectangular waveguide is a common metal waveguide with simple fabrication, low loss and dual polarization. It is often used in antenna feeders requiring dual polarization mode, and also widely used in various resonators and wavelength meters. This paper attempts to design, simulate and optimize rectangular waveguide. By setting the length, width and height of rectangular waveguide, the required rectangular waveguide is designed. The resonant frequency of rectangular waveguide is setting to 9.25 GHz. Moreover, the average gain of radiation direction is required to be as large as possible. After satisfying the requirements, sequential non-linear programming method and genetic algorithm are used to optimize both voltage standing wave ratio and normalized impedance matching. Simulation results show that the proposed method is able to accomplish the optimal design of rectangular waveguide.
Authors: Meijiao Lin (Tianjin Normal University), Xin Zhang (Tianjin Normal University), Yang Li (Tianjin Normal University), Zhou Wu (Chongqing University),
Hide Authors & Abstract

Show Authors & Abstract
12:15 - 12:30
Noise Reduction in Network Embedding

Network Embedding aims to learn latent representations and effectively preserves structure of network and information of vertexes. Recently, embedding networks with rich side information such as vertex’s label and links between vertices has attracted significant interest due to its wide applications such as node classification and link prediction. Current semi-supervised graph embedding algorithms assume the vertex label is ground-truth. However, an important issue is, in real world applications, network always contains mislabeled vertices and edge, which will cause the embedding preserve mistake information. While manually relabel all mislabeled vertex is inapplicable, how to effective reduce noise so as maximize the graph analysis task performance is of great importance. In this paper, we focus on reducing label noise ratio in dataset to obtain more reasonable embedding. We proposed two methods for any semi- supervised network embedding algorithm to tackle it: first, we use a model to identify potential noise vertices and correct it, second, we use two voting strategy to precisely relabel vertex. To the best of our knowledge, we are the first to tackle this issue in network embedding. Our experiments are conducted on three public data sets
Authors: Cong Li (Nanjing University of Aeronautics and Astronautics), Donghai Guan (Nanjing University), Zhiyuan Cui (Nanjing University of Aeronautics and Astronautics), Weiwei Yuan (Nanjing University of Aeronautics and Astronautics), Asad Masood Khattak (Zayed University), Muhammad Fahim (Innopolis University),
Hide Authors & Abstract

Show Authors & Abstract
12:30 - 12:45
Indoor localization Based on Centroid Constraints of AP Quadrilateral Networks

With the development of wireless communication technology at home and abroad, wireless indoor positioning technology has become an indispensable technology. In the current indoor scenes, due to the extensive deployment of Wireless Local Area Network (WLAN) infrastructure, indoor WLAN location method has become a research hotspot. There are a variety of WLAN Access Points (APs) in indoor environment, which can be used in indoor localization. While existing indoor positioning technology often works along with high cost and low system stability especially in complex indoor environment. To solve this problem, a centroid constraint indoor location method based on AP quadrilateral network is proposed. By collecting and processing the Received Signal Strength (RSS) of APs, this method exploits propagation models and trilateration method to get the target quadrilateral centroid set, which can be applied to obtain the ulti-mate estimate coordinates of pending points under the condition of geometric constraints. This method enhances the robustness of indoor localization system and implements low-cost indoor localization.
Authors: Li Xinyue, Zhou Mu (School of Communication and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing), Zhang Zhenya (School of Communication and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing), Liu Zhu (School of Communication and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing),
Hide Authors & Abstract

Show Authors & Abstract
12:45 - 13:00
Partially Overlapping Channel Selection in Jamming Environment: A Hierarchical Learning Approach

This paper solves the channel selection with anti-jamming problem using partially overlapping channel (POC) in limited spectrum environment. Since it is dicult for users to obtain global information of networks, this paper realizes the coordination of channel access by the local information interaction. The channel selection with anti-jamming problem is formulated as a Stackelberg game where the jammer acts as leader and users act as followers. We prove that the game model exists at least one Stackelberg equilibrium (SE) solution. To achieve the equilib- rium, a hierarchical learning algorithm (HLA) is proposed. Based on the proposed method, the system can achieve the improvement of throughput performance by minimizing local interference. Simulation results show the proposed algorithm can achieve good performance under jamming environment, and the network throughput can maintain a stable state with the jamming intensity increasing.
Authors: Lei Zhao (College of Communications Engineering, PLA Army Engineering University, Nanjing, China , Key Embedded Technology and Intelligent System Laboratory, Guilin University of Technology, Guilin, China), Jincheng Ge (Unit 95965 of PLA, Hengshui, China), Kailing Yao (College of Communications Engineering, PLA Army Engineering University, Nanjing, China , Key Embedded Technology and Intelligent System Laboratory, Guilin University of Technology, Guilin, China), Yifan Xu (College of Communications Engineering, PLA Army Engineering University, Nanjing, China , Key Embedded Technology and Intelligent System Laboratory, Guilin University of Technology, Guilin, China), Xiaobo Zhang (College of Communications Engineering, PLA Army Engineering University, Nanjing, China), Menglan Fan (Unit 31102 of PLA, Nanjing, China),
Hide Authors & Abstract

Show Authors & Abstract
13:00 - 13:15
A Q-Learning-based Channel Selection and Data Scheduling Approach for High-frequency Communications in Jamming Environment

The existence of jammer and the limited buffer space bring major challenge to data transmission efficiency in high-frequency (HF) commuication. The data transmission problem of how to select transmission strategy with multi-channel and different buffer states to maximize the system throughput is studied in this paper. We model the data transmission problem as a Makov decision process (MDP). Then, a modified Q-learning with additional value is proposed to help transmitter to learn the appropriate strategy and improve the system throughput. The simulation results show the proposed Q-learning algorithm can converge to the optimal Q value. Simultaneously, the QL algorithm compared with the sensing algorithm has better system throughput and less packet loss.
Authors: Wen Li (College of Communications Engineering, Army Engineering University of PLA, Nanjing, 210000, China), Yuhua Xu (College of Communications Engineering, Army Engineering University of PLA, Nanjing, 210000, China), Qiuju Guo (PLA 75836 Troops, Guangzhou, 510000, China), Yuli Zhang (National Innovation Institute of Defense Technology, Academy of Military Sciences PLA China, Beijing, 100000, China), Dianxiong Liu (College of Communications Engineering, Army Engineering University of PLA, Nanjing, 210000, China), Yangyang Li (College of Communications Engineering, Army Engineering University of PLA, Nanjing, 210000, China), Wei Bai (College of Communications Engineering, Army Engineering University of PLA, Nanjing, 210000, China),
Hide Authors & Abstract

Show Authors & Abstract
13:15 - 13:30
RFID Indoor Location Based on Optimized Generalized Regression Neural Network

Nowadays, location-based services are widely used in social life. Traditional GPS location can provide high-quality location services in outdoor environments, but it is insufficient for indoor location. There are many indoor location technologies, such as WIFI, Bluetooth, ultrasound and RFID. RFID location technology is favored by researchers because of its low cost and high precision. Most of the existing RFID location algorithms are based on RSSI (Received Signal Strength Indicator) measurement. The inaccurate estimation of the path loss parameter may lead to large error when converting RSSI to distance. In order to reduce the error, we propose a new RFID location algorithm. Specifically, the RSSI of the target tag is read in different directions of the antenna, and the position information is predicted by the general regression neural network optimized by the optimization algorithm. The experimental results show the efficiency of our proposed algorithm.
Authors: fangjin chen (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), xiangmao chang (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), xiaoxiang xu (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), Yanjun Lu (Urban Construction College, Wuchang University of Technology),
Hide Authors & Abstract

Show Authors & Abstract
13:30 - 13:45
Beacon in the Air: Optimizing Data Delivery for Wireless Energy Powered UAVs

UAV-aided Internet of Things (IoT) systems enable IoT devices to relay up-to-date information to base stations with UAVs, which extends the IoT network coverage and improves data transmission efficiency. To achieve a perpetual UAV data delivery system, simultaneous wireless data and power transfer (SWIPT) is employed for energy-constrained UAVs to harvest energy from wireless chargers to support data sensing and transmission from IoT devices (e.g., sensors) deployed at different locations. In this paper, the design objective is to pursue the optimal energy charging policy for each UAV considering the system states of location, the queue length and energy storage. We formulate and solve a Markov decision process for the UAV data delivery to optimally take the actions of energy charging, and data delivery to base stations. The performance evaluation shows that the proposed MDP scheme outperforms baseline schemes in terms of lower expected overall cost and high energy efficiency.
Authors: Huajian Jin (Wuhan University of Technology), Jiangming Jin (TuSimple), Yang Zhang (Wuhan University of Technology),
Hide Authors & Abstract

Show Authors & Abstract
13:45 - 14:00
Power Optimization in Wireless Powered based Mobile Edge Computing

Mobile edge computing (MEC) can meet the requirements of high-bandwidth and low delay commanded by the boost developing of mobile network and shorten the network load. This paper investigates a wireless powered MEC system con-sists a single antenna AP, and two single antenna mobile devices which are pow-ered by wireless power transmissions (WPT) from AP. In order to settle the us-ers’ near-far influence, the system will let the mobile devices closer from the AP mobile devices as a relay for unloading. The Objective of this paper is to mini-mize the transmission energy of the AP, taking into account the restraints of the computing task. Our solution is divided into two steps: first, the mathematical model of the prob-lem is listed, and then the optimal solution of each feasible scheme is discussed in a classified manner, and the minimum transmission power of AP is obtained through comparison. Simulation results show that collaboration can reduce ener-gy consumption and improve the user performance.
Authors: Xiaohan Xu (College of Information Engineering, Zhejiang University of Technology, Hangzhou 310023, China), Qibin Ye (College of Information Engineering, Zhejiang University of Technology, Hangzhou 310023, China), Weidang Lu (College of Information Engineering, Zhejiang University of Technology, Hangzhou 310023, China), Hong Peng (College of Information Engineering, Zhejiang University of Technology, Hangzhou 310023, China), Bo Li (School of Information and Electrical Engineering, Harbin Institute of Technology, Weihai 264209, China),
Hide Authors & Abstract

Show Authors & Abstract

Coffee Break 13:40 - 14:10

(Out of Meeting Room)

Session 2.2 (Session Chair: Chenguang Shi) 14:10 - 15:50

Smart Internet of Things
14:15 - 14:30
A smart wearable device for preventing indoor electric shock hazards

The emerging wearable IoT technology has evolved dramatically for the past five years and resulted in a wide adoption in various applications around the world. Electric shock hazard is one of the major indoor hazards for consumers or pets who may fail to recognize the potential electrical hazards. This paper proposes a smart wearable IoT device with risk assessment algorithms for preventing indoor electrical shock hazards. This device consists of two hardware components: the detector integrates a Wi-Fi module, a current sensor, a NFC module, and an Arduino mini module that communicates with a software application to monitor the status of the power switches and connected appliances; the receiver is a passive NFC tag that can be designed as accessories or clothing that people or pets may wear. Based on a set of predefined inference rules, the risk assessment algorithm is capable of evaluating the risks of power switches and appliances. The software application can provide early warnings to the unknowing users where potential electrical shock hazards can be suggested. This paper describes the implementation details as well as the algorithms. Experimental results demonstrate that the proposed smart wearable device can be effective in predicting electric shock hazards in an indoor environment.
Authors: Zaipeng Xie (Hohai University), Hanxiang Liu (Hohai University), Junpeng Zhang (Hohai University), Xiaorui Zhu (Hohai University), Hongyu Lin (Southeast University),
Hide Authors & Abstract

Show Authors & Abstract
14:30 - 14:45
A VLP approach based on a single LED lamp

Visible light positioning (VLP) has become a promising indoor localization approach as it can provide sub-meter localization. VLP usually requires no less than three LED lamps for angle-of-arrival (AoA) or received-signal-strength (RSS) based localization. However, it is hard to identify multiple LED lamps in some indoor environments, such as a long corridor or tunnel, which makes existing VLP useless. To address this problem, we propose a VLP approach using only a single LED lamp. In this approach, we utilize the inertial measurements to infer rotation angles and exploit visual projection geometry to calibrate rotation angles. The experiment results demonstrate the proposed VLP approach can achieve sub-meter positioning accuracy by using only one LED lamp.
Authors: Jing Chen (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), Jie Hao (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), Ran Wang (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), Ao Shen (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), Ze Yu (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics),
Hide Authors & Abstract

Show Authors & Abstract
14:45 - 15:00
A Method of Calculating the Semantic Similarity between English and Chinese Concepts

In the big data era, data and information processing is a common concern of diverse fields. Efficiency and intelligence are two keys of the processing process. To achieve the two keys, searching, defining and building the potential links among heterogeneous data is necessary. Focusing on this issue, this paper proposes a knowledge-driven method to calculate the semantic similarity between (bilingual English-Chinese) words concerning the definitions. This method is built on the knowledge base “HowNet”, which defines and maintains the “atom taxonomy tree” and the “semantic dictionary” - a network of knowledge system describing the relationship between word concepts and the attributes of concepts. The process of semantic similarity analysis is divided into two parts: calculating the atom distance and calculating the definition (concepts) similarity. To improve the efficiency, the non-relational database MongoDB is employed in this method. Taking advantage of MongoDB, such as full indexing and multi-language support, the rich knowledge maintained in HowNet can be fully used. Considering both the structure of HowNet and characteristics of MongoDB, a certain number of equations are defined to calculate the semantic similarity.
Authors: Jingwen CAO (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), Tiexin WANG (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), Wenxin LI (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), Chuanqi TAO (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics),
Hide Authors & Abstract

Show Authors & Abstract
15:00 - 15:15
A Bicycle-borne Sensor Node for Monitoring Air Pollution based on NB-IoT

Nowadays, everybody knows that shared bicycles have become a new type of green transportation in the city. The impact on people's health caused by air pollutants exposed near roads has become a concern in recent years. This paper introduces a device which consists of a particulate detector, temperature, and humidity sensor, micro-SD card, GPS receiver and the NB-IoT communication module. The device can be installed as a sensor node on a shared bike, and a mobile sensor network has been set up on a shared bike to monitor air quality throughout the city, which is of great significance for urban air quality testing.
Authors: Shu Shen (Nanjing University of Posts and Telecommunications), Caixia Lv (Nanjing University of Posts and Telecommunications), Xindi Xu (Nanjing University of Posts and Telecommunications), Xiaoyu Liu (Nanjing University of Posts and Telecommunications),
Hide Authors & Abstract

Show Authors & Abstract
15:15 - 15:30
Power Beacons Deployment in Wireless Powered Communication with Truncated Poisson Cluster Process

Wireless powered communication (WPC) is able to provide the wireless devices (WDs) practically infinite energy for information transmission, by deploying multiple power beacons (PBs) as dedicated energy source. The performance of wireless energy transfer and wireless information transmission may significantly vary depending on the locations of the wireless nodes and the channels used for signal transmission. To enable efficient transmission and overcome the doubly-near-far problem in WPC, this paper proposes a PB deployment strategy where the distribution of PBs is subject to a truncated Poisson cluster process (PCP), and analytically investigates the performance of WPC in terms of the SNR outage probability. Specifically, we consider a harvest-then-transmit communication network, where WDs use RF harvested energy for the information transmission in each time block. The wireless energy transfer between WDs and PBs is achieved by either directed mode (WD is served by the closed PB) or isotropic mode (WD is served by multiple PBs). We first investigate the distribution of the distance between WD and an arbitrary PB in the associated cluster. Then, we derive a numerically computable form of the SNR outage probability for directed mode and a tight upper bound of the SNR outage probability for isotropic mode. Finally, numerical results verify the accuracy of the analytical results, present performance comparisons, and reveal the advantages of the truncated PCP based PB deployment.
Authors: Siyuan Zhou (College of Computer and Information, Hohai University), Jinghang Zhao (College of Computer and Information, Hohai University), Guoping Tan (College of Computer and Information, Hohai University), Xujie Li (College of Computer and Information, Hohai University), Qin Yan (College of Computer and Information, Hohai University),
Hide Authors & Abstract

Show Authors & Abstract
15:30 - 15:45
Mobile Edge Computing-Enabled Resource Allocation for Ultra-Reliable and Low-Latency Communications

Mission critical services and applications with computation intensive tasks require extremely low latency, while task offloading for mobile edge computing (MEC) incurs extra latency. In this work, the optimization of power consumption and delay are studied under ultra reliable and low latency (URLLC) framework in a multiuser MEC scenario. Delay and reliability are relying on users’ task queue lengths, which is attested by probabilistic constraints. Different from the current literature, we consider a comprehensive system model taking into account the effects of bandwidth, computation capability, and transmit power. By introducing the approach of Lyapunov stochastic optimization, the problem is solved by splitting the multi-objective optimization problem into three single optimization problems. Performance analysis is conducted for the proposed algorithm, which illustrates that the tradeoff parameter indicates the tradeoff between power and delay. Simulation results are presented to validate the theoretical analysis of the impact of various parameters and demonstrate the effectiveness of the proposed approach.
Authors: Yun Yu (College of Computer and Information), Siyuan Zhou (College of Computer and Information), Xiaocan Lian (College of Computer and Information), Guoping Tan (College of Computer and Information), Yingchi Mao (College of Computer and Information),
Hide Authors & Abstract

Show Authors & Abstract
15:45 - 16:00
A Bayesian Method for Link Prediction with Considering Path Information

Predicting the links among nodes in a network is an interesting and practical problem. Many methods have been proposed for link prediction separately in network science and computer science. There is a need to combine these two types of methods to further improve the prediction performance. In line with this direction, we study the link prediction problem based on the Bayesian method, with fully considering the information of paths between nodes, and propose a new link prediction method, i.e., path-based Bayesian (PB) method. In this prediction method, we use the Bayesian analysis to deduce the weights of paths of different lengths, and then further consider the contribution of each path of a specific length. Simulation results on real-world networks show that our prediction method has higher prediction accuracy than the mainstream methods.
Authors: Suyuan Zhang (Nanjing University of Science and Technology), Lunbo Li (Nanjing University of Science and Technology), Cunlai Pu (Nanjing University of Science and Technology), Siyuan Zhou (Hohai University),
Hide Authors & Abstract

Show Authors & Abstract
16:00 - 16:15
Identifying sources of random walk-based epidemic spreading in networks

Identifying the sources of epidemic spreading is of critical importance to epidemic control and network immunization. However, the task of source identification is very challenging, since in real situations the dynamics of the spreading process is usually not clear. In this paper, we formulate the multiple source epidemic spreading process as the multiple random walks, which is a theoretical model applicable to various spreading processes. Considering the different influence of distinct epidemic sources on the observed infection graph, we derive the maximum likelihood estimator of the multiple source identification problem. Simulation results on real-world networks and network models, such as the Price model and Erd\"os-R\'enyi (ER) model, demonstrate the efficiency of our estimator. Furthermore, we find that the efficiency of our estimator increases with the enhancement of network sparsity and heterogeneity.
Authors: Bo Qin (Nanjing University of Science and Technology), Cunlai Pu (Nanjing University of Science and Technology),
Hide Authors & Abstract

Show Authors & Abstract
Room #4

Lunch 10:00 - 11:30

(Buffet)
Room #5

Dinner 16:30 - 19:30

Conference Banquet & Best Paper Award Ceremony
Day 2 25/08/2019
Room #1

Session 3.1 (Session Chair: Tiexin Wang) 06:30 - 08:30

Machine Learning I
06:30 - 06:45
Active Sampling Based on MMD for Model Adaptation

In this paper, we demonstrate a method to do transfer learning with minimal supervised information. Recently, researchers have proposed various algorithms to solve transfer learning especially unsupervised domain adaptation. They mainly focused on how to learn a good common representation and used it directly in the downstream task. Unfortunately, they neglect the fact that such representation may not capture target-specific feature for target task. In order to solve this problem, this paper attempts to utlize labled data in target daomain to capture target-specific feature. Now it's a challenge that how to seek as few supervised information as possible to achieve a good result . To overcome this challenge, we actively select instance for training and model adaptation based on MMD criterion. During this process,we tries to label some valuable target data to capture targt-specific featrue and fintune the classcifier networks. We choose a batch of data in target domain far from common represention space and having maximum entropy. The first requirement is helpful to learn a good representation for target domain and the second requirement tries adjust the classifier performance. Eventually, we experiment our method on several datasets which show significant improvements and competitive advantages against common methods.
Authors: Qi Zhang (Nanjing University of Aeronautics and Astronautics), Donghai Guan (Nanjing University), Weiwei Yuan (Nanjing University of Aeronautics and Astronautics), Asad Masood Khattak (Zayed University),
Hide Authors & Abstract

Show Authors & Abstract
06:45 - 07:00
Optimal Dwell Time for Frequency Hopping in a Stackelberg Game with a Smart Jammer

Frequency hopping (FH) technique is usually used to anti-jamming communication. Frequency dwell time is an important parameter for FH com-munication. Short dwell time will reduce the communication efficiency due to frequency switching time, while long dwell time will increase the time to be jammed after the sensing of a smart jammer. The dwell time of the cognitive user and the sensing time of the jammer are interactive. We formulate the interactions between the user and the jammer as a Stackelberg game. The jammer first senses the user’s operating frequency and then jams the user based on the sensing result. The user determines its dwell time according to the reward under the jamming. A tiered reinforcement learning algorithm is proposed to solve the game. The optimal dwell time of the user is given when the Stackelberg Equilibrium is achieved.
Authors: Long Yu (Sixty-third Research Institute, National University of Defense Technology), Yonggang Zhu (Sixty-third Research Institute, National University of Defense Technology), Yusheng Li (Sixty-third Research Institute, National University of Defense Technology),
Hide Authors & Abstract

Show Authors & Abstract
07:00 - 07:15
An Active Noise Correction Graph Embedding Method Based on Active Learning for Graph Noisy Data

In various scenarios of the real world, there are various graph structures. Most graph structures have the characteristics of huge structure and large space overhead. Graph embedding is an effective method to solve this problem, which converts graph structure information into low-dimensional dense real vector and maps it to a low-dimensional latent space. In the real world, label acquisition is expensive, and there may be noise in the data. It is very important to select the points that need expert marking in the labeled nodes to maximally find the noise points that have the greatest impact on the graph structure and improve the performance of the node classification. In this paper, we propose an active corrective graph embedding method (ANCGE) based on active learning for graph noisy data. Given the label budget, we use semi-supervised graph embedding algorithm to find more noise in labeled nodes by active learning. And when the active corrective graph embedding method chooses the noise nodes which need to be corrected, the noise nodes which have large amount of representativeness and great influence on the graph are selected as far as possible by considering the structure of the graph. The experimental results on three open datasets demonstrate the effectiveness of our method and its stability under different noise rates.
Authors: Zhiyuan Cui (Nanjing University of Aeronautics and Astronautics), Donghai Guan (Nanjing University), Cong Li (Nanjing University of Aeronautics and Astronautics), Weiwei Yuan (Nanjing University of Aeronautics and Astronautics), Asad Masood Khattak (College of Technological Innovation),
Hide Authors & Abstract

Show Authors & Abstract
07:15 - 07:30
Semi-supervised Learning via adaptive low-rank graph

Graph-based semi-supervised learning (SSL) is one of the most popular topics in the past decades. Most conventional graph-based SSL methods utilize two stage-approach to infer the class labels of the unlabeled data, i.e. it firstly constructs a graph for capturing the geometry of data manifold and then perform SSL for prediction. However, the graph construction and SSL stages are separate. They do not share common information to enhance the performance of classification. In addition, the graph should also be adaptive to the parameters and datasets. In this paper, we aim to handle the above issues. To achieve adaptiveness of SSL, we adopt a bilinear low-rank model for graph construction, where the coefficient matrix of the low-rank model is calculated through an adaptive and efficient procedure the corresponding constructed graph can capture the global structure of data manifold. Meriting from such a graph, we then propose a unified framework for scalable SSL, where we have involved the graph construction and SSL into a unified optimization problem. As a result, the discriminative information learned by SSL can be provided to improve the discriminative ability of graph construction, while the updated graph can further enhance the classification results of SSL. Simulation indicates that the proposed method can achieve better classification and clustering performance compared with other state-of-the-art graph-based SSL methods.
Authors: Mingbo Zhao, Zhang Jiang (Donghua University), Cuili Yang (Beijing University of Technology),
Hide Authors & Abstract

Show Authors & Abstract
07:30 - 07:45
Transaction Cost Analysis via Label-Spreading Learning

When the investment institution analyzes the transaction cost of stock orders, it is costly to obtain the transaction cost of the stock orders by trading it, in contrast, many simulated trading orders cannot get the exact transaction cost. Due to lack of enough labeled data, it is usually hard to estimate accurate transaction cost of stock orders using a supervised learner. Label-spreading, a graph-based semi-supervised learner, can integrate a small number of labeled real orders and a large number of unlabeled simulated orders, and train a learner simultaneously. Using a RBF kernel, the learner constructs a graph structure through the spatial similarity measure between the transaction cost samples, and propagates the label through edges of graph in high-dimensional space. The results of experiments show that the label-spreading learner can make full use of the information of unlabeled data to improve classification of transaction cost.
Authors: Pangjing Wu (College of Computer and Information, Hohai University, Nanjing, China.), Xiaodong Li (College of Computer and Information, Hohai University, Nanjing, China.),
Hide Authors & Abstract

Show Authors & Abstract
07:45 - 08:00
A Novel PCA-DBN based Bearing Fault Diagnosis Approach

This paper is concerned with fault diagnosis problem of a widely used component in vast rotating machinery, rolling element bearing. We propose a novel intelligent fault diagnosis approach based on principal component analysis (PCA) and deep belief network (DBN) techniques. By adopting PCA technique, the dimension of raw bearing vibration signals is reduced and the bearing fault features are extracted in terms of primary eigenvalues and eigenvectors. Parts of the modi ed samples are trained by DBN for fault classi cation and diagnosis and the rest are tested to examine the algorithm. A distinctive feature of this approach is that it requires no complex signal processing procedure of bearing vibration signals. The experimental results demonstrate the effectiveness of the PCA-DBN based fault diagnosis approach with a more than 90% accuracy rate.
Authors: Jing Zhu (Nanjing University of Aeronautics and Astronautics), Tianzhen Hu (Nanjing University of Aeronautics and Astronautics),
Hide Authors & Abstract

Show Authors & Abstract
08:00 - 08:15
A Visual Semantic Relations Detecting Method Based on WordNet

WordNet is an excellent relational dictionary maintaining huge number of words, words’ senses and semantic relations among the words. Many research works, which concern the data and information processing problem, have been carried out on the basis of WordNet to achieve automatic semantic relations detection purpose. However, considering the index-based text storage structure of WordNet, which lacks native deep semantic topology, it cannot support complex semantic relations detection or semantic reasoning directly. Focusing on this issue, this paper proposes a visual semantic relations detecting method (VSRDM) based on WordNet. To enhance the efficiency, the object-oriented graph database Neo4j is combined. The main contribution of this work is “implementing the transformation between the relational storage structures of WordNet to the ternary storage structure of Neo4j”. Taking advantage of the graph searching algorithms and visual displays of Neo4j, the efficiency and per-formance of VSRDM is guaranteed. Certain potential usages of VSRDM, such as employing as semantic dictionary, providing reasoning auxiliary, are discussed at the end of this paper.
Authors: Wenxin LI (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), Tiexin WANG (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), Jingwen CAO (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), Chuanqi TAO (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics),
Hide Authors & Abstract

Show Authors & Abstract
08:15 - 08:30
Statement generation based on big data for keyword search

Natural language generation(NLG) is the process of automatically generating a high-quality natural language text through a planning process based on some key information. Regular NLG generates sentences by analyzing grammatical and semantics, generating rules, and then organizing elements based on rules and heuristics. However, sentences generated by such methods are too strict, poorly scalable and difficult to adapt to the changing language style of human beings nowadays. Our goal is to generate smooth, personal, multi-sentence text for end users. This paper introduces a new NLG system, which can generate distinctive statements, and discard the knowledge of semantics, syntax etc,which are required by the original rule-based generation statements. This system turns out to be simple and efficient. We obtain required corpus from the network, and then use the idea of the search engine to find sentences from a large amount of data that matches the meaning of the keyword provided by users. Such generated sentences are more consistent with people's daily life. Finally, we apply our system in the web commentary domain, evaluating our system based on three criteria. The result show that our system works well in this field and can continue to deepen.
Authors: Qingqing Liu (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics), Zhengyou Xia (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics),
Hide Authors & Abstract

Show Authors & Abstract

Coffee Break 08:10 - 08:40

(Out of Meeting Room)

Session 4.1 (Session Chair: Ken Cai) 08:40 - 10:20

Application of Machine Learning I
08:45 - 09:00
Knoweldge graph embedding Based on Hyperplane and Quantitative Credibility

Knowledge representation learning is one of the research hotspots in the field of knowledge graph in recent years. How to improve the training algorithm of knowledge representation model and improve the accuracy of knowledge graph knowledge completion prediction is the main research goal in this field. Applying more implicit semantic information to model training is the primary means of im-proving accuracy. The traditional method does not consider the change of knowledge validity with time. So for this problem, we study the distribution law of quantitative changes of knowledge, design the model to simulate the quantita-tive changes of knowledge, put forward the concept of quantitative credibility, and apply it to the training algorithm of the model, and put forward A new learn-ing method of knowledge representation QCHyTE. We compare the trained mod-el with the best-recognized algorithms, and the results show that our improved algorithm greatly improves the prediction accuracy of the model.
Authors: Shuo Chen (State Grid Liaoning Electric Power Supply Co), Lin Qiao (State Grid Liaoning Electric Power Supply Co), Biqi Liu (State Grid Liaoning Electric Power Supply Co), Jue Bo (State Grid Liaoning Electric Power Supply Co), Yuanning Cui (Nanjing University of Aeronautics and Astronautics), Jing Li (Nanjing University of Aeronautics and Astronautics),
Hide Authors & Abstract

Show Authors & Abstract
09:00 - 09:15
Artificial Intelligence Approaches for Urban Water Demand Forecasting: A Review

In various fields such as medicine, marketing, engineering military e.t.c. Artificial intelligence approaches are been applied mainly due to their powerful symbolic reasoning capability, flexibility, modeling and forecasting power. In this paper an attempt to review urban water demand using various artificial based approaches such as fuzzy logic systems, support vector machines, extreme learning machines, ANN, ARIMA as well as hybrid models which consist of a combination of two or more AI models for urban water demand forecasting are applied. The paper illustrates how the different AI approaches plays a major role in urban water demand while recommending some future work directions.
Authors: Abdullahi Muhammad (Hohai University), Xiaodong Li (Hohai University), Jun Feng (Hohai University),
Hide Authors & Abstract

Show Authors & Abstract
09:15 - 09:30
Cyberbullying Detection with BiRNN and Attention Mechanism

While the social network has brought a lot of conveniences to our lives, it has also caused a series of severe problems, which include cyberbullying. Cyberbullying is an aggressive and intentional act carried out by a group or an individual to attack a victim on the Internet. Most of the existing works related to cyberbullying detection focus on making use of swearwords to classify text or images with short titles. Although previous methods such as SVM and logistic regression show some advantages in the accuracy of detection, few of them capture the semantic information of non-swearwords which could also make big difference to the final results. In this paper, we propose to use BiRNN and attention mechanism to identify bullies. BiRNN is used to integrate the contextual information, and the attention model reflects the weight of different words for classification. Meanwhile, we convert the severity calculated by the attention layer to the level of cyberbullying. Experiments conducted on three real-world text datasets show that our proposed method outperforms the state-of-art algorithms on text classification and identification effect.
Authors: ANMAN ZHANG (Nanjing University of Aeronautics and Astronautics), BOHAN LI (Nanjing University of Aeronautics and Astronautics), SHUO WAN (Nanjing University of Aeronautics and Astronautics), KAI WANG (Nanjing University of Aeronautics and Astronautics),
Hide Authors & Abstract

Show Authors & Abstract
09:30 - 09:45
Sewage treatment control method based on Genetic-SOFNN

Aiming at the problem that the dissolved oxygen concentration in wastewater treatment is difficult to control accurately, this paper proposes a self-organizing fuzzy neural network controller based on genetic process. In the neural network structure adjustment stage, the fuzzy rules of the controller and the parameters of the neurons are initialized by genetic process, and the deleted rules with the information of the neurons be retained which are merged to improve the performance of the self-organizing mechanism. At the same time, the projection gradient method is used to train the controller parameters to improve the control precision of the controller. Experiments show that the fuzzy neural network control method proposed in this paper can accurately control the concentration of dissolved oxygen in the sewage treatment process. Compared with other control methods, the control accuracy is improved and the stable operation of the sewage treatment process is guaranteed.
Authors: zhuang yang (Faculty of Information Technology, Beijing University of Technology), cuili yang (Faculty of Information Technology, Beijing University of Technology), junfei qiao (Faculty of Information Technology, Beijing University of Technology),
Hide Authors & Abstract

Show Authors & Abstract
09:45 - 10:00
Latent Flow Patterns Discovery by Dockless Bicycle Trajectory Data for Demand Analysis

The dockless shared bikes have flourished as a new concept that has emerged in recent years. Due to shared bikes allow users to find, pick-up and drop-off bikes anywhere via GPS-based smartphone application, we can extract bicycle flow patterns from bicycle trip data for better urban planning and Point-of-Interest(POI) recommendation. In order to solve these problems, in this paper, through rich bicycle activity spatio-temporal representations in bicycle trip logs, we design a graph clustering model with sparsity constraints that combine time information to identify potential patterns of bicycle flow. By using the characteristic typologies of bicycle flow patterns from the model, we then compare historical trip logs and POI information with the flow patterns and thus give suggestions for further urban planning and investment optimization. Furthermore, our experiments via Mobike trajectory data demonstrate the effectiveness of bicycle flow pattern discovery.
Authors: Chao Ling (Nanjing University of Aeronautics and Astronautics), Jingjing Gu (Nanjing University of Aeronautics and Astronautics), Ming Sun (Nanjing University of Aeronautics and Astronautics),
Hide Authors & Abstract

Show Authors & Abstract
10:00 - 10:15
Sniffer Patrolling for Wireless Traffic Monitoring via Deep Reinforcement Learning

Without introducing interference to the legacy wireless networks, passive traffic monitoring can be used for net- work diagnosis and radio frequency management in spectrum- sharing wireless networks by deploying wireless sniffers to monitor traffic on different channels. This motivates the sniffer- channel assignment problem, i.e., assigning each wireless sniffer a proper operating channels with the aim of tracking the target signals or data packets. The existing approaches for sniffer- channel assignment were usually designed for the scenarios where the behaviors of malicious or suspect wireless users are known. In this paper, we focus on a cognitive radio network without knowledge about the users’ behavior, in which the wireless sniffers can be deployed to patrol different locations and to meet certain performance requirement. Due to the dynamics of network environment and huge state space, we propose a novel deep reinforcement learning (DRL) approach to determine the patrolling route for each sniffer. Via numerical simulations, we show that the proposed DRL approach can achieve better performance than that of the conventional Deep Q-network.
Authors: Zhipeng Chen (School of Electronic Information and Communications, Huazhong University of Science and Technology, China), Yutong Xie (Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, China), Jing Xu (School of Electronic Information and Communications, Huazhong University of Science and Technology, China),
Hide Authors & Abstract

Show Authors & Abstract
10:15 - 10:30
Travel Time Estimation and Urban Key Routes Analysis Based on Call Detail Records Data: A Case Study of Guangzhou City

Nowadays, the study of urban traffic characteristics has become a major part of city management. With the popularization of the mobile communication network, the call detail records(CDR) data has become an important resource for the study of urban traffic, containing abundant temporal and spatial information of urban population. To excavate the traffic characteristics of Guangzhou city, this paper focuses on two aspects: travel time estimation and the analysis of key routes in urban area. First, we propose a method of estimating urban travel time based on traffic zones division using the traffic semantic attributes. According to the users' flow features extracted from CDR, we determine the traffic semantic attributes of the areas covered by base stations. With these semantic attributes, we cluster the cell areas into several traffic zones using a K-means method with a weighting dissimilarity measure. Then travel time between different positions in Guangzhou city is estimated using the key locations of traffic zones, with an accuracy rate of 67%. Furthermore, we depict the key routes of Guangzhou city utilizing a DBSCAN method with the users' trajectories extracted from the CDR data. The results of acquired routes are validated by actual traffic conditions, also providing some extra discoveries. Our works illustrate the effectiveness of CDR data in urban traffic and provide ideas for further research.
Authors: Weimin Mai (School of Electronics and Information Technology, Sun Yat-sen University, Guangzhou 510006, China), Shaohang Xie (School of Electronics and Information Technology, Sun Yat-sen University, Guangzhou 510006, China), Xiang Chen (School of Electronics and Information Technology, Sun Yat-sen University, Guangzhou 510006, China),
Hide Authors & Abstract

Show Authors & Abstract
10:30 - 10:45
Task Allocation in Multi-agent Systems Using Many-objective Evolutionary Algorithm NSGA-III

Task allocation is an important issue in multi-agent systems, and finding the optimal solution of task allocation has been demonstrated to be an NP-hard problem. In many scenarios, agents are equipped with not only communication resources but also computing resources, so that tasks can be allocated and executed more efficiently in a distributed and parallel manner. Presently, many methods have been proposed for distributed task allocation in multi-agent systems. Most of them are either based on complete/full search or local search, and the former usually can find the optimal solutions but requires high computational cost and communication cost; the latter is usually more efficient but could not guarantee the solution quality. Evolutionary algorithm (EA) is a promising optimization algorithm which could be more efficient than the full search algorithms and might have better search ability than the local search algorithms, but it is rarely applied to distributed task allocation in multi-agent systems. In this paper, we propose a distributed task allocation method based on EA. We choose the many-objective EA called NSGA-III to optimize four objectives in task allocation simultaneously. Experimental results show the effectiveness of the proposed method, and compared with the full search strategy, the proposed method could solve task allocation problems with more agents and tasks.
Authors: Jing Zhou (Dalian University of Technology), Xiaozhe Zhao (Dalian University of Technology), Dongdong Zhao (Wuhan University of Technology), Zhong Lin (Dalian Naval Academy),
Hide Authors & Abstract

Show Authors & Abstract
Room #2

Session 3.2 (Session Chair: Congduan Li) 06:30 - 08:10

Machine Learning II
06:30 - 06:45
Batch gradient training method with smoothing l0 regularization for echo state networks

This paper considers the batch gradient method with the smoothing l0 regularization (BGSL0) for training and pruning feedforward neural networks. We show why BGSL0 can produce sparse weights, which are crucial for pruning networks. We prove both the weak convergence and strong convergence of BGSL0 under mild conditions. The decreasing monotonicity of the error functions during the training process is also obtained. Two examples are given to substantiate the theoretical analysis and to show the better sparsity of BGSL0 than three typical lp regularization methods.
Authors: Ahmad Zohaib (Faculty of Information Technology, Beijing University of Technology), Kaizhe Nie (Beijing Unversity of Technology), junfei qiao (Faculty of Information Technology, Beijing University of Technology), Cuili Yang (Faculty of Information Technology, Beijing University of Technology),
Hide Authors & Abstract

Show Authors & Abstract
06:45 - 07:00
A backward learning algorithm for feature selection in polynomial echo state networks

Recently a polynomial echo state network (PESN), an extension of the original ESN, was proposed and its polynomial output weights were augmented by the high order statistics of full input features. However, there may be noisy and redundant features to construct the polynomial function as the output weights, which result in the high computational complexity and degrade the estimation accuracy. To remove insignificant features and to reduce computational complexity, a backward learning algorithm for feature selection in polynomial ESN (BS-PESN) is proposed. Firstly, the criterion of removing feature is utilized to delete the insignificant features one by one from PESN. Then, an iterative strategy is also used to reduce the training computational burden in BS-PESN. Finally, 10 UCI regression datasets experiments are made which show that the proposed approach can have better prediction accuracy and less testing time than the original PESN.
Authors: Cuili Yang (Faculty of Information Technology, Beijing University of Technology), Xinxin Zhu (Faculty of Information Technology, Beijing University of Technology), Junfei Qiao (Faculty of Information Technology, Beijing University of Technology),
Hide Authors & Abstract

Show Authors & Abstract
07:00 - 07:15
Using LSTM, GRU and Hybrid Models for Streamflow Forecasting

Forecasting streamflow discharge of rivers have economic impact as well as reducing the effects of flood in flood prone regimes by giving early warning signs. To reduce the effects of flood in those regimes, a powerful class of machine learning models called long short-term memory (LSTM) and Gated recurrent units (GRU) models, which have become popular and applied in time series forecasting because they are explicitly designed to avoid the long-term dependency problems was applied. LSTM and GRU have also demonstrated their capacity in sequence modelling, speech recognition, streamflow forecasting e.t.c. In this paper we proposed a hybrid model for streamflow forecasting using 35 consecutive years Model Parameter Estimation Experiment (MOPEX) data set of ten different basins having different basin characteristics from different climatic regions in United States and applied it in streamflow forecasting. Finally, the proposed hybrid model performance is compared with the conventional LSTM and GRU models. Our experiments on 10 MOPEX datasets demonstrate that, although the the proposed hybrid model outperforms conventional LSTM with regards to streamflow forecasting but the performance is almost same with GRU and is therefore highly recommended as an efficient tool in hydrological fields for streamflow forecasting.
Authors: Abdullahi Muhammad (Hohai University), Xiaodong Li (Hohai University), Jun Feng (Hohai University),
Hide Authors & Abstract

Show Authors & Abstract
07:15 - 07:30
Backscatter-aided Hybrid Data Offloading in Mobile Edge Computing via Deep Reinforcement Learning

Data offloading in mobile edge computing (MEC) allows the low power IoT devices in the edge to optionally offload power-consuming computation tasks to MEC servers. In this paper, we consider a novel backscatter-aided hybrid data offloading scheme to further reduce the power consumption in data transmission. In particular, each device has a dual-mode radio that can offload data via either the conventional active RF communications or the passive backscatter communications with extreme low power consumption. The flexibility in the radio mode switching makes it more complicated to design the optimal offloading strategy, especially in a dynamic network with time-varying workload and energy supply at each device. Hence, we propose the deep reinforcement learning (DRL) framework to handle huge state space under uncertain network state information. By a simple quantization scheme, we design the learning policy in the Double Deep Q-Network (DDQN) framework, which is shown to have better stability and convergence properties. The numerical results demonstrate that the proposed DRL approach can learn and converge to the maximal energy efficiency compared with other baseline approaches.
Authors: Yutong Xie (Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, China), Zhengzhuo Xu (School of Electronic Information and Communications, Huazhong University of Science and Technology, China), Jing Xu (School of Electronic Information and Communications, Huazhong University of Science and Technology, China), Shimin Gong (School of Intelligent Systems Engineering, Sun Yat-sen University, Shenzhen, China), Yi Wang (School of Innovation and Entrepreneurship, Southern University of Science and Technology, China),
Hide Authors & Abstract

Show Authors & Abstract
07:30 - 07:45
An Efficient Federated Learning Scheme with Differential Privacy in Mobile Edge Computing

In this paper, we consider a mobile edge computing (MEC) system that multiple users participate in the federated learning protocol by jointly training a deep neural network (DNN) with their private training datasets. The main challenges of applying federated learning to MEC are: 1) it incurs tremendous computational cost by carrying out the deep neural network training phase on the resource-constraint mobile edge devices; 2) existing literature demonstrates that the parameters of a DNN trained on a dataset can be exploited to partially reconstruct the training samples in original dataset. To address the aforementioned issues, we introduce an efficiently private federated learning scheme in mobile edge computing, named FedMEC, with model partition technique and differential privacy method in this work. The experimental results demonstrate that our proposed FedMEC scheme can achieve high model accuracy under different perturbation strengths.
Authors: Jiale Zhang (Nanjing University of Aeronautics and Astronautics), Junyu Wang (Nanjing University of Aeronautics and Astronautics), Yanchao Zhao (Nanjing University of Aeronautics and Astronautics), Bing Chen (Nanjing University of Aeronautics and Astronautics),
Hide Authors & Abstract

Show Authors & Abstract
07:45 - 08:00
Joint Power and Channel Selection for Anti-jamming Communications: A Reinforcement Learning Approach

In this paper, the decision-making problem for anti-jamming communications is studied. Most of the existing anti-jamming researches mainly focus on the single-domain anti-jamming such as power domain or frequency domain, which has limited performance facing strong jamming. Therefore, to effectively deal with some jamming attack, this paper proposes a multi-domain joint anti-jamming scheme, and considers the power domain and the frequency domain jointly. By modeling the anti-jamming process as a Markov decision process (MDP), reinforcement learning (RL) is adopted to solve the MDP. Then, the multi-domain joint anti-jamming algorithm is proposed to find the optimal decision-making strategy. Moreover, the proposed algorithm is verified to converge to an effective strategy. Simulation results show that the proposed algorithm has better throughput performance than the sensing-based random selection algorithm.
Authors: Xufang Pei (College of Communications Engineering, Army Engineering University of PLA), Ximing Wang (College of Communications Engineering, Army Engineering University of PLA), Lang Ruan (College of Communications Engineering, Army Engineering University of PLA), Luying Huang (College of Communications Engineering, Army Engineering University of PLA), Xingyue Yu (College of Communications Engineering, Army Engineering University of PLA), Heyu Luan (College of Communications Engineering, Army Engineering University of PLA),
Hide Authors & Abstract

Show Authors & Abstract
08:00 - 08:15
Motion Classification based on sEMG Signals using Deep Learning

Nowadays, surface electromyography (sEMG) signal plays an important role in helping physically disabled people during daily life. The development of electronic information technology has also led to the emergence of low-cost, multi-channel, wearable sEMG signal acquisition devices. Therefore, this paper proposes a new motion recognition model based on deep learning to improve the accuracy of motion recognition of sEMG signals. The model uses architecture including 6 convolutional layers and 6 pooling layers which enables the Batch Normalization layer to enhance network performance and prevent overfitting. In the experiment, NinaPro DB5 data set was used for training and testing. The data set consists of sEMG signal data collected by the double Myo armbands which contains data of 52 movements for 10 subjects. The results show that the accuracy of about 90% can be achieved when using 52 sEMG signal data from every single subject or all subjects.
Authors: Shu Shen (Nanjing University of Posts and Telecommunications), Kang Gu (Nanjing University of Posts and Telecommunications), Xinrong Chen (Fudan University), Ruchuan Wang (Nanjing University of Posts and Telecommunications),
Hide Authors & Abstract

Show Authors & Abstract
08:15 - 08:30
Realization of Transmission Control Protocol Based on μC/OS-II

As the number of embedded devices has increased dramatically, communication between these devices has also become an important issue. Currently, many open source TCP/IP stacks have been packaged, but they cannot be used directly on embedded devices. Therefore, this paper proposes to transplant the open source TCP/IP protocol stack μC/IP into the embedded device to complete the communi-cation between devices. This paper modifies the transmission control proto-col(TCP) of μC/IP so that embedded devices can communicate based on the TCP. Finally, the experimental test results show that the TCP of the transplanted TCP/IP protocol stack can work smoothly, and can be compatible with standard TCP to complete communication between embedded devices.
Authors: Qianyuan Wang (Tongji University), Yujun Gao (Tongji University),
Hide Authors & Abstract

Show Authors & Abstract

Coffee Break 08:10 - 08:40

(Outing of Meeting Room)

Session 4.2 (Session Chair: Xiangmao Chang) 08:40 - 10:20

Application of Machine Learning II
08:45 - 09:00
Robustness Analysis on Natural Language Processing based AI Q&A Robots

Recently, the natural language processing (NLP) techniques based intelligent question and answering (Q&A) robots have been widely used, not only in general application areas, but also in professional business or government applications. However, the robustness and security issues of these NLP based artificial intelligence (AI) Q&A robots have not been studied yet. In this paper, we analysis the robustness problems in current Q&A robots, which includes four aspects: 1) semantic slot settings are incomplete; 2) sensitive words are not filtered efficiently and completely; 3) Q&A robots return the search results directly; 4) unsatisfactory matching algorithms and inappropriate matching threshold setting. Then, we design and implement two types of tests, bad language and user’s typos, to evaluate the robustness of several state-of-the-art Q&A robots. Experiment results show that these common inputs (bad language and user’s typos) can successfully make these Q&A robots malfunction, denial of service, or speaking dirty words. Besides, we also propose possible countermeasures to enhance the robustness of these Q&A robots. To the best of authors’ knowledge, this is the first work on analyzing the robustness problems of intelligent Q&A robots. This work can hopefully help improve the robustness and accuracy of the Q&A robots.
Authors: Chengxiang Yuan (Nanjing University of Aeronautics and Astronautics), Mingfu Xue (Nanjing University of Aeronautics and Astronautics), Lingling Zhang (Nanjing University of Aeronautics and Astronautics), Heyi Wu (Southeast University),
Hide Authors & Abstract

Show Authors & Abstract
09:00 - 09:15
BMI-Matching: Map-Matching with Bearing Meta-Information

Map-matching is a fundamental preprocessing step for many applications which aligns a trajectory represented by a sequence of sampling points with the city road network on a digital map. The past years have witnessed the dramatic advance of location-acquisition technologies. With the help of GPS-embedded devices tracking moving objects, a lot of GPS trajectories can be collected. However, the raw positions cap- tured by GPS devices usually can not reflect the real positions because of physical constrictions. So it becomes significant to accurately match GPS trajectories to road network. In practice, noisy and low sampling rate data are big challenges for map-matching problem. Unfortunately, most existing methods which only consider the position of the object or the topology structures of the road network may not be able to handle. We propose a method called BMI-matching(map-matching with bearing meta-information) which not only considers the two factors above but also focuses on the moving object bearing. Based on bearing, we can calculate the direction similarity between moving object and road segments to determine selecting which road segment is appropriate. We conduct experiments on real dataset and compare our method with two state- of-the-art algorithms. The results show that our approach gets better performance on accuracy.
Authors: Dawei Wang (Nanjing University of Aeronautics and Astronautics), JingJing Gu (Nanjing University of Aeronautics and Astronautics),
Hide Authors & Abstract

Show Authors & Abstract
09:15 - 09:30
Using Speech Emotion Recognition to Preclude Campus Bullying

Campus bullying could have extremely adverse impact on pupils,leading to physical harm,mental disease,or even ultra behaviour like suicide.Hence,an accurate and efficient anti-bullying approach is badly needed.A campus bullying detection system based on speech emotion recognition is proposed in this paper to distinguish bullying situations from nonbullying situations.Initially,a Finland emotional speech database is divided into two parts, namely training-data and testing-data,from which MFCC(Mel Frequency Cepstrum Coeffi- cient) parameters are garnered. Subsequently, ReliefF feature selection algorithm is applied to select the useful features to form a matrix.Then its dimensions is diminished with PCA(Principle Component Analysis) algorithm.Finally, KNN(K-Nearest Neighbor) algorithm is utilized to train the model. The final simulations show a recognition rate of 80.25%, verifying that this model is able to provide a useful tool for bullying detection.
Authors: Jianting Guo (Harbin Institute of Technology at Weihai), Haiyan Yu (Harbin Institute of Technology at Weihai),
Hide Authors & Abstract

Show Authors & Abstract
09:30 - 09:45
Optimal Control of Navigation Systems with Time Delays Using Neural Networks

In this paper, an online adaptive dynamic programming (ADP) scheme is proposed to achieve the optimal regulation control of nonlinear navigation time-delay system with control constraints in virtue of Lyapunov stability theo-ries and neural networks (NNs) techniques.We firstly investigate the stability on nonlinear navigation time delay systems concerning input constraints. By means of liner matrix inequalities (LMIs), we construct an appropriate optimal control scheme. Then we utilize an NN to estimate the performance function and the constrained to approximate the optimal control policy such that the NN weight can be online tuned. The NN weight estimate error is proved convergent to zero. Finally, numerical example is presented to illustrate our results.
Authors: Jing Zhu (Nanjing University of Aeronautics and Astronautics), Yijing Hou (Nanjing University of Aeronautics and Astronautics),
Hide Authors & Abstract

Show Authors & Abstract
09:45 - 10:00
Multi-Spectral Palmprint Recognition with Deep Multi-View Representation Learning

With the increasing application of biometrics in security systems, palmprint recognition technology, as an emerging biometric technology, has received more and more attention in recent years. Building a palmprint recognition system usually involves image acquisition, preprocessing, feature extraction and matching. Feature extraction and matching are usually the most essential processes in palmprint recognition, and researchers with different backgrounds like biometrics, pattern recognition, computer vision, and neural networks have invested a lot of attention. In this paper, we propose a deep multi-view representation learning based multi-spectral palmprint fusion method, which uses deep neural networks to extract feature representation of multi-spectral palmprint images for palmprint classification. In this manner, the unique features of different spectral palmprint images can be used to learn a view-invariant representation of each palmprint. By using view-invariant representation, we can get better palmprint recognition performance than single modality. Experiments are performed on multi-spectral palmprint database to validate the effectiveness of the proposed method.
Authors: Xiangyu Xu (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, 3 Nanjing 211106, China), Nuoya Xu (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, 3 Nanjing 211106, China), Huijie Li (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, 3 Nanjing 211106, China), Qi Zhu (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, 3 Nanjing 211106, China),
Hide Authors & Abstract

Show Authors & Abstract
10:00 - 10:15
Reinforcement Learning for HEVC Screen Content Intra Coding on Heterogeneous Mobile Devices

Intra coding of HEVC screen content coding has to evaluate HEVC intra coding modes and additional modes for screen contents, which poses a challenge for coding such a content on mobile devices. Furthermore, the heterogeneous mobile devices have varying complexity requirements. In this paper, a flexible screen content intra coding scheme is proposed, which can trade between encoding complexity and rate-distortion performance degradation via reinforcement learning (RL). Through the design of states, actions, and more importantly, the reward function for RL, the proposed scheme can learn a flexible coding policy offline. Experimental results show the effectiveness of the proposed scheme.
Authors: Yuanyuan Xu (Hohai University), Quanping Zeng (Hohai University),
Hide Authors & Abstract

Show Authors & Abstract
10:15 - 10:30
An Accelerated PSO based self-organizing RBF neural network for nonlinear system identification and modeling

In this paper, an accelerated particle swarm optimization (APSO) based radial basis function neural network (RBFNN) is designed for nonlinear system modeling. In APSO-RBFNN, the center, width of hidden neurons, weights of output layer and network size are optimized by using the APSO method. Two nonlinear system modeling experiments are used to illustrate the effectiveness of the proposed method. The simulation results show that the proposed method has obtained good performance in terms of network size and estimation accuracy.
Authors: Ahmad Zohaib (Faculty of Information Technology, Beijing University of Technology), Cuili Yang (Faculty of Information Technology, Beijing University of Technology), Junfei Qiao (Faculty of Information Technology, Beijing University of Technology),
Hide Authors & Abstract

Show Authors & Abstract
10:30 - 10:45
Non-negative Matrix Factorization with Community Kernel for Dynamic Community Detection

Finding community structures from user data has become a hot topic in network analysis. However, there are rarely effective algorithms about the dynamic community detection. To recalculate the whole previous nodes and deal with excessive calculation is used to solve the problem of dynamically adding community nodes in the previous researches. In this paper, we propose an incremental community detection algorithm without recalculating the whole previous nodes using the incremental non-negative matrix factorization (INMF). In this algorithm, community kernel nodes with the largest node degree and adjacent triangle ratio is selected to calculate the data feature matrix, then the complexity of the calculations is largely simplified by reducing the dimension of the data feature matrix. We also propose a strategy to solve the problem of ensuring the feature space dimension and community number of NMF. We discuss our method with several previous ones on real data, and the results show that our method is effective and accurate in find potential communities.
Authors: Saisai Liu (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing, 210016, China), Zhengyou Xia (College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing, 210016, China),
Hide Authors & Abstract

Show Authors & Abstract
Room #3

Lunch 10:00 - 11:30

(Buffet)