学　术

　　为搭建我校博士后学术交流平台，促进学术水平提升，学校博士后管理办公室组织开展博士后学术沙龙活动。本次沙龙由我校博士后刘畅、郭化盐、萧洒、蒯小燕和何振清分享其研究成果，诚挚邀请感兴趣的师生参加。

　　一、时　间：2018年5月31日（周四）13：00

　　二、地　点：清水河校区经管楼宾诺咖啡

　　三、主办单位：电子科技大学博士后管理办公室

　　　　承办单位：通信抗干扰技术国家级重点实验室　　电子科技大学博士后联谊会

　　四、活动安排：

　　报告一：

　　（1）主　题：Deep CNN: A Versatile Feature Exploration for Smart Spectrum Sensing in Cognitive Radio Networks

　　（2）主讲人：刘　畅  通信抗干扰技术国家级重点实验室博士后 

　　（3）交流内容：

　　Deep Learning (DL) has been applied for many fields, such as computer vision and natural language processing, given its expressive capacity and convenient optimization capability. Recently, the potential application of DL to communications has been increasingly recognized. In DL, deep convolution neural network (CNN) is a good solution to the Layer-Wise-Pre-training procedure for the multilayer neural network. By imposing the convolution operation on layers, the CNN is able to discover interesting structure in the signal, especially the essential feature of the raw communications data.

　　This report will first give a brief overview of the deep CNN. Based on the fundamentals of deep CNN, this report will analyze the spectrum sensing model in cognitive radio networks (CRNs). To address the detection problem under complicated environment, we design a data-driven test statistic using the classical LeNet-5 network. Since CNN can capture image feature very well, we use sample covariance matrix as input image of CNN to exploit the inherent feature of received samples, and propose a novel LeNet-Sensing (LS) detection algorithm. Through the offline training and online deployment, the proposed LS detector can achieve outstanding detection performance, which is very close to the optimal theoretical bound. Therefore, this work gives the perspective on how to use CNN in CRNs, which inspires new introspection of the future communication systems.

　　主讲人简介：

　　Chang Liu received the B.S. degree in electronic information engineering from Dalian Maritime University, Dalian, China, in 2012, and the Ph.D. degree in signal and information processing, Dalian University of Technology, China. From 2015 to 2016, he was a visiting scholar in department of electrical engineering and computer science at University of Tennessee, Knoxville, USA. He is currently a postdoctoral research fellow with the National Key Laboratory of Science and Technology on Communications, UESTC. His research interests include deep learning, spectrum sensing in cognitive radio, and statistical signal processing.

　　报告二：

　　（1）主　题：Cognitive IoT: Next Generation Cellular IoT

　　（2）主讲人：郭化盐  通信抗干扰技术国家级重点实验室博士后 

　　（3）交流内容：

　　In this talk, we briefly introduce the cognitive IoT, which is an emerging technology for the next generation Cellular IoT. As the third wave of information technology after internet and mobile communications, the IoT has become a term that encompasses a wide range of new applications and services. Two of the primary challenges for the practical deployment of large numbers of IoT sensor nodes include limited available spectrum and battery lifetime. To realize ultra low power consumption transmissions, backscatter communication (BackCom) has emerged as a promising passive radio technique, in which the backscatter devices (BD) do not need for power-hungry radio-frequency (RF) components such as up-converters and power amplifiers. However, the ultra low transmit power also results in low spectrum efficient, and thus it is not practical to allocate large amount of dedicated spectrum for backscatter transmission. Cognitive IoT is an emerging technology to overcome above limitation, which is an integration of the well-known cognitive radio concept and the BackCom technology. In cognitive IoT system, the backscatter system shares not only the same spectrum, but also the same RF source with the legacy system (e.g., TV, cellular, or Wi-Fi systems). The BD transmits data to the reader over the received modulated'' legacy RF signals by intentionally varying the antenna impedances. Owing to the spectrum sharing nature, the backscatter system achieves better spectrum utilization efficiency than the traditional BackCom, while it still inherits the ultra low power consumption property for transmission. As a newborn technology, the study of cognitive IoT is still at its nascent stage, and there are various first order technical challenges arising from the data communication and networking perspectives. This talk will briefly introduce the recent research results about cognitive IoT by our team, mainly containing information theoretical studies and high throughput transceiver designs.

　　主讲人简介：

　　H.-Y. Guo received the B.S. degree in electronic information science and technology from the Beijing University of Posts and Telecommunications, and the Ph.D. degree from School of Electronics Engineering and Computer Science, Peking University. He is now a post-doc majored in cognitive radio and intelligence communications.

　　报告三：

　　（1）主题：Robust resource allocation for full-duplex enabled Cognitive radio networks

　　（2）主讲人：萧 洒  通信抗干扰技术国家级重点实验室博士后 

　　（3）交流内容：

　　This talk will introduce a novel algorithm for resource allocation in full-duplex enabled cognitive networks, where the channel state information of the links between secondary users (SUs) and primary users (PUs) is uncertain. To protect the transmission of the PUs from interference generated by the SUs, robust optimization theory is utilized to characterize the channel uncertainty and a resource allocation problem is formulated by jointly optimizing sub-channel assignment, user pairing, and power allocation. By using the dual method, the original resource allocation problem is decomposed into a primal problem and a dual problem. The concave-convex procedure is adopted to transform the primal problem into a tractable form through sequential convex approximations while the sub-gradient method is used to solve the dual problem. Simulation results demonstrate the effectiveness of the proposed algorithm.

　　主讲人简介：

　　Sa Xiao received Ph.D. degrees from the University of Electronic Science and Technology of China, Chengdu, China, in 2017. From February 2015 to August 2015, he was a visiting student with the Department of Electrical and Computer Engineering, Southern Illinois University, Carbondale, IL, USA. He also worked as a visiting student in Division of Electrical and Computer Engineering, Louisiana State University, Baton Rouge, LA, USA, from August 2015 to February 2017. He is currently a postdoctoral research fellow with the National Key Laboratory of Science and Technology on Communications, UESTC. His research interests include resource allocation in device-to-device communications, full-duplex communications, and heterogeneous networks.

　　报告四：

　　（1）主　题：Robust Low-Rank Matrix Factorization

　　（2）主讲人：何振清  通信抗干扰技术国家级重点实验室博士后

　　（3）交流内容：

　　Low-rank matrix factorization (LRMF) aims at approximating a data matrix, possibly involving an incomplete observation, by the product of two low-rank factor matrices. This enables an intrinsic low dimensional representation and unravels hidden feature information of the input high dimensional data. There have been numerous applications where the problem of LRMF appears, for example, structure from motion, collaborative filtering, face recognition, graph clustering, and social networks, to name a few. In this talk, we concentrate on the problem of (asymmetric and symmetric) LRMF in the presence of outliers, resulting from some abnormal behaviours, such as malicious tampering, amplifier saturation and sensor failures. We address this problem under a formulation involving some robust loss functions, instead of the standard squared-error loss. Specifically, by leveraging the block majorization-minimization principle, we develop a computationally efficient block iteratively reweighted algorithmic framework, with closed-form descent calculation per iteration and provable convergence guarantee toward a stationary point of the original intractable problem. Numerical experiments using both synthetic data and real data demonstrate the superiority of the proposed algorithm.

　　主讲人简介：

　　Zhen-Qing received the Ph.D. degree in Communications and Information System from the University of Science and Technology of China (UESTC) in 2017. He currently works as a Postdoctoral Fellow in the National Key Laboratory of Science and Technology on Communications of UESTC, under the supervision of Prof. Xiaojun Yuan. His general research interests lie in the areas of signal processing, machine learning, and wireless communication.

　　报告五：

　　（1）主　题：Message-Passing Based OFDM Receiver for Time-Varying Multipath Channels

　　（2）主讲人：蒯小燕  通信抗干扰技术国家重点实验室博士后

　　（3）交流内容：

　　Orthogonal frequency division multiplexing (OFDM) has been extensively studied in high-rate transmission over frequency-selective fading channels. Time-varying channels arise in many practical scenarios, such as underwater acoustic communication. However, the orthogonality of the OFDM system is compromised in time-varying channels, where the subcarriers suffer from inter-carrier interference (ICI) due to the Doppler effect. Joint channel estimation and signal detection can improve the receiver performance. In this talk, we concentrate on the receiver design for time-varying OFDM system. We develop a novel joint channel estimation and data detection receiver which employs a turbo message passing framework for efficient ICI suppression. Specifically, a forward and backward message passing algorithm is adopted for data detection; Channel estimation can be realized by using the generalized approximate message passing (GAMP) algorithm.

　　主讲人简介：

　　Xiaoyan Kuai obtained the Ph. D. degree in Communications and Information Engineering from Xiamen University in 2017. She currently works as a Postdoctoral Fellow in the National Key Laboratory of Science and Technology on Communications of UESTC, under the supervision of Prof. Xiaojun Yuan. Her general research interests lie in the areas of wireless communications, underwater acoustic communications, and signal processing.

　　　　　　　　　　　　　　　　　　　电子科技大学博士后管理办公室

　　　　　　　　　　　　　　　　　　　　　　　　2018年5月28日