contentsmeeting.xidian.edu.cn/.../20120610/20120610134501.pdf · and optimization from academia,...

Contents

Xidian University, July of 2012 1

Organizers........................................................................ 2

Topics and Scope ............................................................. 3

Pre-worshop Tutorials...................................................... 4

Worshop Program ............................................................ 6

Speakers & Abstracts

Part I: Tutorials ......................................................... 11

Part II: Talks ............................................................. 15

Map of Xidian University.............................................. 44

Organizers

Workshop on Mathematical Issues in Information Sciences 2

Workshop Organizers

Hongwei Liu Xidian University

Zhi-Quan (Tom) Luo University of Minnesota

Yu-Guang Shi Peking University

Ming Jiang Peking University

Yu-Hong Dai Chinese Academy of Sciences

Local Organizers

Lin-Rang Zhang Xidian University

Jian-Guo Huang Northwestern Polytechnical University

Yuan-Yuan Zuo Xidian University

Kehu Yang Xidian University

Topics And Scope


In recent years there has been a surge of research in the interfaces between mathematical optimization, statistics, information theory, data mining, signal processing and wireless communication, leading to new powerful mathematical techniques as well as efficient algorithms for important applications in information sciences. The interplay between the information sciences and the applied mathematics communities has been greatly beneficial to both sides: modern optimization and statistics have significantly broadened the class of information science problems considered solvable, while applications in data mining, signal processing and communications provide great impetus for new algorithmic tools and analysis in statistics and optimization. These render an international workshop necessary to bring together leading researchers in signal processing, communications and optimization from academia, industry and government, and showcase recent new results on algorithmic methods and complexity analysis in diverse areas such as estimation and detection theory, MIMO radar, interference management, image/speech processing, beamforming and array processing, communication systems and networks, and so on. The 2012 MIIS workshop is for the goal http://see.xidian.edu.cn/workshop/miis2012/ and offers a platform for international researchers to communicate with the leading experts and each other.

The 2012 MIIS workshop will feature invited presentations/panel discussions in a single track by twenty three leading experts, July 9-13,2012, covering four topics: Sparse Signal Processing

Optimization Methods,

Signal Processing Aplications

Wireless Communications

Three pre-workshop tutorials will be provided to the interested graduate students

and young researchers, July 7-8,2012, focusing on important mathematical techniques in information sciences: Semidefinite Relaxation and Its Applications in SP/Comm.

Variational Inequalities: A Framework for Multiuser Comm/SP

Algorithms for Sparse Optimization

A poster session will offer an opportunity for workshop participants to present

their latest research in information sciences, July 9 - 13, 2012. Two best posters will be selected and announced during the workshop banquet. Each winner will be offered a certificate and a cash award of RMB500.

Saturday, July 7, 15:00–18:00


Tutorial Session #1 Time: Saturday, July 7, 15:00 – 18:00 Place: Building J - 112, Main Campus of Xidian University Session Chair: Jianguo Huang, Northwestern Polytechnical University Speaker:

Ken Ma, Chinese University of Hong Kong, Hong Kong Title:

Semidefinite Relaxation and Its Applications in Signal Processing and Communications

Sunday, July 8, 9:00 – 18:00


Tutorial Session #2 Time: Sunday, July 8, 9:00 – 12:00 Place: Building J - 112, Main Campus of Xidian University Session Chair: Kehu Yang, Xidian University Speaker:

Daniel Palomar, Hong University of Sciences and Technology, Hong Kong

Title:

Variational Inequality Theory: A Mathematical Framework for Multiuser Communication Systems and Signal Processing

Tutorial Session #3 Time: Sunday, July 8, 15:00 – 18:00 Place: Building J - 112, Main Campus of Xidian University Session Chair: Guangming Shi, Xidian University Speaker:

Wotao Yin, Rice University, USA

Title:


Workshop Program-DAY 1


Monday, July 9, 7:00–18:20 7:00 – 8:30 Registration

8:30 – 9:00 Opening Session Session Chair: Hongwei Liu Opening Speeches by Baoyan Duan, Zhaotian Zhang, Yuguang Shi, Yuhong Dai, Zhi-Quan (Tom) Luo Session Topic: Sparse Signal Processing Time: Monday, July 9, 9:00 – 17:20 Place: Building J - 112, Main Campus of Xidian University Morning Session Chair: Zhi-Quan Luo, University of Minnesota

9:00 – 10:00 Georgios Giannakis, University of Minnesota In-network Rank Minimization and Sparsity Regularization: Algorithms and Application to Unveiling Traffic Anomalies

10:00 – 10:30 Coffee Break & Poster Session 10:30 – 11:30 Michael Elad, Technion, Isreal

Sparse and Redundant Representation Modeling of Images: Theory and Applications

11:30 – 12:30 Wotao Yin, Rice University Global Linear Convergence of the Alternating Direction Method and the Linearized Bregman Method

12:30 – 14:30 Lunch Break & Poster Session Afternoon Session Chair: Michael Elad, Technion

14:30 – 15:30 Zhengyuan Xu, Tsinghua University Perturbation Analysis for Geolocation Performance with Biased Range Measurements

15:30 – 16:30 Jiansheng Yang, Peking University High Order Total Variation Minimization Method for Interior Tomography

16:30 – 17:20 Coffee Break & Poster Session 17:20 – 18:20 Panel on Sparse Signal Processing Panel Chair: Yi Wang, Cornell University Panelists: Michael Elad (Technion), Georgios Giannakis (Minnesota) Hongwei Liu (Xidian University), Wotao Yin (Rice University) Xinbo Gao (Xidian University) 18:40 – 20:30 Reception

DAY 2-Workshop Program


Tuesday, July 10, 8:45 – 18:20 Session Topic: Optimization Methods Time: Tuesday, July 10, 8:45 – 16:30 Place: Building J - 112, Main Campus of Xidian University Morning Session Chair: Jong-Shi Pang, UIUC

8:45 – 9:45 Stephen Boyd, Stanford University

Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers

9:45 – 10:45 Zhi-Quan Luo, University of Minnesota Analysis of first order optimization methods for large scale optimization

10:45 – 11:30 Coffee Break & Poster Session 11:30 – 12:30 Zongben Xu, Xi’an Jiaotong University

Sparse Information Processing: Non-Convex Regularization Theory with Applications

12:30 – 14:30 Lunch Break & Poster Session Afternoon Session Chair: Stephen Boyd, Stanford University

14:30 – 15:30 Yinyu Ye, Stanford University A Unified Framework for Dynamic Information Market Design: A Case of On-Line Convex Optimization

15:30 – 16:30 Yu-Hong Dai, Chinese Academy of Sciences On the Subspace Minimization Conjugate Gradient Method

16:30 – 17:20 Coffee Break & Poster Session 17:20 – 18:20 Panel on Research Funding in Information Sciences Panel Chair: Zongben Xu, Xi’an Jiaotong University Panelists: Zhaotian Zhang (NSF, China), Zhi Tian (NSF, USA), Maria Greco (University of Pissa, Italy), Yu-Guang Shi (Peking University, China)

19:00 – Banquet and Dance/Music Show



Wednesday, July 11, 8:45 – 12:30 Session Topic: Signal Processing Applications Time: Wednesday, July 11, 8:45 – 12:30 Place: Building J - 112, Main Campus of Xidian University Morning Session Chair: Wei Yu, University of Toronto

8:45 – 9:45 Shuguang Cui, University of Texas A&M

Asymptotic Results for Distributed Estimation with and without Node Failures

9:45 – 10:45 Daniel P. Palomar, HKUST, Hong Kong The Ubiquitous Variational Inequality Theory: From Wireless Comm. to Smart Grids and Financial Systems

10:45 – 11:30 Coffee Break & Poster Session

11:30 – 12:30 Yan Zhu, Peking University Secure Computation Overview and Research Progress

12:30 – Free Discussion & Visit to the Terra Cotta Museum

DAY 4-Workshop Program


Thursday, July 12, 8:45 – 18:20 Session Topic: Wireless Communications Time: Thursday, July 12, 8:45 – 16:30 Place: Building J - 112, Main Campus of Xidian University Morning Session Chair: Georgios Giannakis, University of Minnesota

8:45 – 9:45 Jong-Shi Pang, UIUC

Joint Sensing and Power Allocation in Cognitive Radio Games 9:45 – 10:45 Wei Yu, University of Toronto

Optimization of Cooperative Wireless Multicell Networks

10:45 – 11:30 Coffee Break & Poster Session 11:30 – 12:30 Wing-Kin Ma, Chinese University of Hong Kong

Physical-Layer Multicasting for MISO Downlink: Going beyond Transmit Beamforming

12:30 – 14:30 Lunch Break & Poster Session

Afternoon Session Chair: Shuguang Cui, University of Texas A&M

14:30 – 15:30 Joakim Jaldén, KTH, Sweden

Algorithms on random lattices in wireless communications

15:30 – 16:30 Martin Schubert, Technical University of Berlin Interference Calculus -- A General Framework for Interference Management and Network Utility Optimization

16:30 – 17:20 Coffee Break & Poster Session 17:20 – 18:20 Panel on Wireless Communications Panel Chair: Jiandong Li, Xidian University Panelists: Shuguang Cui (Texas A&M), Joakim Jalden (KTH), Martin Schubert (Technical University of Berlin) Yi Zhan (JiangNan Electron. Commun. Res. Inst.), Jian-Guo Huang (Northwestern Polytechnical University)



Friday, July 13, 8:45 – 17:20 Session Topic: Signal Processing Applications Time: Friday, July 13, 8:45 – 16:30 Place: Building J - 112, Main Campus of Xidian University Morning Session Chair: Hongwei Liu, Xidian University

8:45 – 9:45 Jingdong Chen, Northwestern Polytechnic University

Multichannel Acoustic Signal Processing for Spatial Sound Acquisition and Delivery

9:45 – 10:45 Maria Greco, University of Pissa Coherent Radar Target Detection in Heavy-Tailed Compound-Gaussian Clutter

10:45 – 11:30 Coffee Break & Poster Session 11:30 – 12:30 Ming Jiang, Peking University

Phase Unwrapping Problem 12:30 – 14:30 Lunch Break & Poster Session Afternoon Session Chair: Daniel Palomar, HKUST

14:30 – 15:30 Xiang-Gen Xia, University of Delaware

Quantitative SNR Studies for Joint Time-Frequency Analysis with Applications in SAR/ISAR Imaging

15:30 – 16:30 Wenchang Sun, Nankai University Average Sampling and Signal Denoising

16:30 – 17:20 Panel on Emerging Techniques in Radar Applications Panel Chair: Hongwei Liu, Xidian University Panelists: Jianqi Wu (Chair of Chinese Radar Society), Maria Greco (University of Pissa), Jia Xu (Tsinghua University), Tao Zeng (Beijing Inst. of Technology), Jianyu Yang (Univ. of Electron. Sci. and Tech. of China)， Lin Ma (Chief Scientist, China Electronics Technology Group Corperation) 17:20 – 17:40 Presentation of Poster Awards and Closing



Part I: Tutorials



Semidefinite Relaxation and Its Applications in Signal

Processing and Communications

Wing-Kin (Ken) Ma Department of Electronic Engineering, The Chinese University of Hong Kong

Semidefinite relaxation (SDR) has recently been recognized as a key technique in signal processing and communications. It is a powerful approximation technique for a host of difficult optimization problems, notably nonconvex quadratically constrained quadratic programs. SDR has found numerous applications; among them particularly important applications are MIMO detection (which covers multi-user, multi-antenna, space-time,...), transmit beamforming (which covers classical single-cell multiuser downlinks, multicell coordinated multiuser downlinks, unicasting and multicasting, cognitive radio, physical layer security, one-way and two-way relays...), and sensor network localization. And the application scope is still expanding. This talk aims at giving an overview of SDR. I will describe essential ideas and practical deployment aspects of SDR, and summarize some of the key theoretical results offered by optimization researchers. And certainly, the talk will cover the important SDR applications, and some very forefront advances in those applications. Notes:

15:00-18:00, July 7(Tutorial)


Variational Inequality Theory: A Mathematical

Framework for Multiuser Communication Systems

and Signal Processing

Daniel P. Palomar Department of Electronic and Computer Engineering,

Hong Kong University of Science and Technology The use of convex optimization is ubiquitous in communications and signal processing in the design of single-user and multiuser systems with efficient practical algorithms. Other systems require a different formulation based on the so-called complementary problems. Yet, many other communications and networking problems composed of multiple agents with individual interests are better formulated within the realm of game theory, which is a field of applied mathematics that describes and analyzes scenarios with interactive decisions. The mathematical tools offered by convex optimization theory, complementary problems, and classical game theory may not be satisfactory to properly study general Nash Equilibrium problems with arbitrary objective functions and strategy sets and to devise distributed algorithms along with their convergence properties. A more general framework suitable for investigating and solving various optimization problems and equilibrium models even when classical game theory may fail, is known to be the Variation Inequality (VI) problem that constitutes a very general class of problems in nonlinear analysis and has been developed in the operation research literature. VI theory is very promising, but so far it has not been properly exploited in practical applications arising in engineering. There is only a handful of papers in the signal processing and wireless communication communities that use VI to solve practical problems. The few existing applications of VI theory in signal processing and communications are just the tip of the iceberg and it is precisely the aim of this tutorial is to take VI a step further. In particular, the aim is twofold: 1) to provide a unified framework based on VI theory (for solving arbitrary equilibrium problems with distributed algorithms), and 2) to apply such a framework to an array of different applications in the areas of signal processing, communications, networking, and even finance engineering. Notes:

(Tutorial) 9:00-12:00, July 8



Wotao Yin Department of Computational and Applied Mathematics, Rice University

This tutorial will give an overview of the modern algorithms for sparse optimization. It includes (i) prox-linear, dual, primal-dual, and projection-based algorithms for L1 and L1-like optimization, as well as their variants for problems with special structured data and solutions, (ii) algorithms for joint/group sparse optimization, (iii) algorithms based on non-convex objective functions and/or support detection, (iv) algorithms for low-rank matrix completion and reconstruction, and (v) decentralized sparse optimization algorithms. Notes:

15:00-18:00, July 8(Tutorial)


Part II: Talks

9:00-10:00, July 9


In-network Rank Minimization and Sparsity

Regularization: Algorithms and Application to

Unveiling Traffic Anomalies

Georgios B. Giannakis University of Minnesota

Given a limited number of entries from the superposition of a low-rank matrix plus the product of a known fat compression matrix times a sparse matrix, recovery of the low-rank and sparse components is a fundamental task subsuming compressed sensing, matrix completion, and principal components pursuit. This talk presents algorithms for distributed sparsity-regularized rank minimization over networks, when the nuclear- and l1-norms are used as surrogates to the rank and nonzero entry counts of the sought matrices, respectively. While nuclear-norm minimization has well-documented merits when centralized processing is viable, non-separability of the singular-value sum challenges its distributed minimization. To overcome this limitation, an alternative characterization of the nuclear norm is advocated which leads to a separable, yet non-convex cost minimized via the alternating-direction method of multipliers. The resultant per-node iterations incur affordable complexity and message passing among single-hop neighbors to approach the (centrally attainable) global optimum regardless of initialization. Possible applications presented will include unveiling traffic anomalies in backbone networks, predicting network-wide path latencies, and mapping the RF ambiance using wireless cognitive radios. Notes:

10:30-11:30, July 9


Sparse and Redundant Representation Modeling of

Images: Theory and Applications

Michael Elad Technion, Israel

This talk focuses on the use of sparse and redundant representations and learned dictionaries for image denoising and other related problems. We discuss the the K-SVD algorithm for learning a dictionary that describes the image content efficiently. We then show how to harness this algorithm for image denoising, by working on small patches and forcing sparsity over the trained dictionary. The above is extended to color image denoising and inpainting, video denoising, and facial image compression, leading in all these cases to state of the art results. We conclude with more recent results on the use of several sparse representations for getting better denoising performance. An algorithm to generate such set of representations is developed, and our analysis shows that by this we approximate the minimum-mean-squared-error (MMSE) estimator, thus getting better results. Notes:

11:30-12:30, July 9


Global Linear Convergence of the Alternating

Direction Method and the Linearized Bregman Method

Wotao Yin Department of Computational and Applied Mathematics, Rice University

The talk shows that the alternating direction method (ADM) applied to minimizing f(x) + g(y) subject to Ax + By = b converges linearly in various different cases. One case is that one of f and g is strongly convex and has Lipschitz gradient, and the corresponding matrix A or B has full row rank. The talk also shows that global linear convergence holds when the ADM subproblems solved inexact by one iteration of gradient descent or prox-linear descent. In addition, it is shown that the linearized Bregman method applied to minimizing ||x||_1 + alpha ||x||_2^2 subject to Ax = b has global linear convergence and can return exact solutions for compressive sensing. Notes:

14:30-15:30 July 9


Perturbation Analysis for Geolocation Performance

with Biased Range Measurements

Zhengyuan (Daniel) Xu Department of Electronic Engineering, Tsinghua University

Perturbation analysis is a powerful tool for various applications in practical conditions, such as matrix computation, wireless channel estimation, and communication receiver performance prediction. Jointly with Dr. Ning Liu and Dr. Brian M. Sadler, we study geolocation performance based on random biased range measurements. We develop weighted least squares (WLS) and maximum likelihood (ML) geolocation estimators, and apply a perturbation analysis technique to find bias and mean square error (MSE) expressions for these estimators. The results are functions of the measurement bias and variance, as well as the network geometry. They are applied to optimize sensor placement for improving the overall geolocation accuracy. Numerical examples are presented to verify the analysis and study some cases of interest. Notes:

15:30-16:30, July 9


High Order Total Variation Minimization Method for

Interior Tomography

Jiansheng Yang Peking University

Interior problem (IT) in Computerized Tomography refers to reconstruct the image in a region of interest (ROI) only using the projection data associated with the ROI. The classic wisdom showed that the IT doesn’t have unique solution. The well known total variation (TV) minimization method can be utilized to capture the image of piecewise constant type and single out the image from all feasible ones. We propose a definition of high order total variation (HTV) in order to apply HTV minimization method to capture the image of piecewise polynomial type. Furthermore, we investigate the high order interior problem (that is, to reconstruct the image in the ROI only using the high order differential projection data associated with the ROI) and have the similar results. In this talk, I will present more details about the work. Notes:

8:45-9:45, July 10


Distributed Optimization and Statistical Learning via

the Alternating Direction Method of Multipliers

Stephen Boyd, Neal Parikh, Eric Chu, Borja Peleato, and Jon Eckstein Department of Electrical Engineering, Stanford University

Problems in areas such as machine learning and dynamic optimization on a large network lead to extremely large convex optimization problems, with problem data stored in a decentralized way, and processing elements distributed across a network. We argue that the alternating direction method of multipliers is well suited to such problems. The method was developed in the 1970s, with roots in the 1950s, and is equivalent or closely related to many other algorithms, such as dual decomposition, the method of multipliers, Douglas-Rachford splitting, Spingarn's method of partial inverses, Dykstra's alternating projections, Bregman iterative algorithms for $\ell_1$ problems, proximal methods, and others. After briefly surveying the theory and history of the algorithm, we discuss applications to statistical and machine learning problems such as the lasso and support vector machines, and to dynamic energy management problems arising in the smart grid. Notes:

9:45-10:45, July 10


Analysis of First Order Methods for Large Scale

Optimization Problems

Zhi-Quan Luo

Department of ECE，University of Minnesota A popular approach to solve a large scale optimization problem under independent constraints is to cyclically update a subset of variables by minimizing a locally tight convex upper bound of the original (possibly nonsmooth) cost function. This approach includes the well known block coordinate descent method (BCD), the block coordinate proximal point method (BCD) and the expectation maximization (EM) method, among others. In this work, we establish the convergence of the method under mild assumptions on the convex upper bound used at each iteration. Our work unifies, extends and strengthens the existing convergence analysis of the BCD and the EM method, and can be used to derive the convergence of block successive upper minimization methods for tensor decomposition, linear transceiver design in wireless networks, and DC (difference of convex functions) programming, among others. In addition, we will provide rate of convergence analysis for the alternating direction method of multipliers (ADMOM) in the absence of strong convexity. Our analysis uses the theory of error bounds and is applicable to ADMOM methods for more than two blocks. Notes:

11:30-12:30, July 10


Sparse Information Processing: Non-Convex

Regularization Theory With Applications

Zongben Xu Xi'an Jiaotong University

Sparse information processing (SIP) has attracted a great deal of attention in recent years, which aims to find sparse solution(s) of a representation or an equation. In practice, SIP leads to a constrained optimization problem, that is, minimizing a sparsity measure of the transformed unknown variables subject to some linear or nonlinear constraints. The problem is normally solved through various regularization techniques. When the sparsity measure is taken as the indicator function of a closed convex set, the regularization problem can be solved by the well-known gradient projection method. Also, when the sparsity measure is taken as the lower semi-continuous convex function, the problem then can be effectively solved by the classical Moreau’s forward-backward splitting method. However, the non-convex sparsity measures are advocated in many applications such as compressed sensing, image processing and SAR imaging, and hence, a similar Moreau’s splitting theory is anticipated for the non-convex regularization technique. In this talk, we take the $L_q$ (0<q<1) quasi-norm as an example to show the possibility of such extension of Moreau’s theory. For the special case of q=1/2, we develop an $L_{1/2}$ regularization theory, showing that the developed theory provides a successful practice of generalizing the classical Moreau’s theory to non-convex case. We demonstrate that the extended theory have a wide scope of novel applications, especially in signal processing, image processing and SAR imaging. Notes:

14:30-15:30, July 10


A Unified Framework for Dynamic Information Market

Design: A Case of On-Line Convex Optimization

Yinyu Ye Department of Management Science and Engineering, Stanford University

Recently, several pari-mutuel mechanisms have been introduced to organize prediction markets, such as the logarithmic scoring rule, the cost function formulation, and sequential convex pari-mutuel mechanism (SCPM).In this work, we develop a unified framework that bridges these seemingly unrelated models for centrally organizing contingent-claim markets. Our framework establishes necessary and sufficient conditions for designing mechanisms with many desirable properties such as proper scoring, truthful bidding (in a myopic sense), efficient computation, controllable risk-measure, and guarantees on the worst-case loss. As a result, we develop the very first proper, truthful, risk-controlled, loss-bounded, and polynomial-time scoring rule, which neither of the previous proposed mechanisms possesses simultaneously. Thus, in addition to providing a general framework that unifies and explains all the existing mechanisms, our work would be an effective and instrumental tool in designing new market mechanisms. We also discuss applications of our framework to general open markets for dynamic resource pricing and allocation. Notes:

15:30-16:30, July 10


On the Subspace Minimization Conjugate Gradient

Method

Yu-Hong Dai AMSS, Chinese Academy of Sciences

The linear conjugate gradient method is an optimal method for solving symmetrical and positive definite linear equations. The proposition of limited-memory BFGS method and Barzilai-Borwein gradient method, however, heavily restricted the use of conjugate gradient method in large-scale optimization. This is, to the great extent, due to the requirement of a relatively exact line search at each iteration and the loss of conjugacy property of the search directions in various occasions. In this talk, I shall pay much attention on the subspace minimization conjugate gradient method by Yuan and Stoer (1995). Some nice theoretical properties of the method will be explored and some promising nu-merical results will be provided. Consequently, we can see that the subspace minimization conjugate gradient method can become a strong candidate for large-scale optimization. Notes:

8:45-9:45, July 11


Asymptotic Results for Distributed Estimation with

and without Node Failures

Shuguang Cui Beijing University of Posts and Telecommunications

We study the distributed estimation and detection problem in a sensor network, where each networked sensor converges to the same intelligence just based on local observations and information exchanges with its neighbors. We consider both the cases with and without random node failures. For the case without node failures, we estimate the state of a potentially unstable linear dynamical system in the framework of distributed Kalman filtering. In particular, it is shown that, in a weakly connected communication network, there exist (randomized) gossip based information dissemination schemes leading to a stochastically bounded estimation error at each sensor for any non-zero rate $\bar{\gamma}$ of inter-sensor communication. In particular, the conditional estimation error covariance sequence at each sensor is shown to evolve as a random Riccati equation (RRE) with Markov modulated switching, which is analyzed through a random dynamical system (RDS) formulation. The estimation error at each sensor is shown asymptotically converging to an invariant distribution and the overall performance gets close to the centralized estimator exponentially fast over the inter-node communication rate. For the case with random node failures (sensing unit failures), we consider the problem of distributed estimation of an unknown deterministic scalar parameter. We assume that the observation process at a node chooses randomly from two modes: with mode one corresponding to the desired signal plus noise observation mode (a valid sensor); and mode two corresponding to pure noise with no signal information (an invalid sensor). With no prior information on the local sensing modes (valid or invalid), we introduce a learning-based distributed estimation procedure, the mixed detection-estimation (MDE) algorithm, based on closed-loop interactions between the iterative distributed mode learning and estimation, for which the convergence is established. We also show that in the high signal-to-noise ratio (SNR) regime the MDE estimation error converges to that of an centralized estimator with perfect information about the node sensing modes. Notes:

9:45-10:45, July 11


The Ubiquitous Variational Inequality Theory: From

Wireless Comm. to Smart Grids and Financial

Systems

Daniel P. Palomar Department of Electronic and Computer Engineering,

Hong Kong University of Science and Technology During the last decade, the use of convex optimization has become standard in the communications and signal processing communities for the design of single-user and multiuser systems with efficient practical algorithms. However, convex optimization has its limitations that can be overcome with VI theory. Indeed, more recently, other communications and networking problems, composed of multiple agents with individual interests, have been effectively formulated within the VI framework with very practical algorithms. This talk will show that the potential employment of the VI theory goes beyond wireless communications. In particular, we will overview two recent applications in the very different areas of smart grids and financial systems. Notes:

11:30-12:30, July 11


Secure Computation Overview and Research

Progress

Yan Zhu Peking University

Secure computation is a cryptography technology which enables two or more mutually distrusting parties to collaboratively evaluate functions without revealing their inputs to the other parties. Secure computation is one of important branches of modern cryptography, and the theoretical foundation of construction of cryptographic protocols. Especially with the emergence of cloud computing, secure computation has become an important tool to ensure the security of important business on the untrusted cloud platform. In this presentation, we give an overview of new progress of secure computation from three categories: multi-party, two-party, and single-party (also called fully Homomorphic encryption). We also introduce some new research fields and ideas: secure comparison problem, secure decision problem, Multi-Hop Homomorphic Encryption, and so on. We hope that these researches will contribute to a deeper understanding of secure computing, and also help to understand the core technologies of practical applications, including secure database retrieval, outsourcing data processing and secure identity management mechanisms. Notes:

8:45-9:45, July 12


Joint Sensing and Power Allocation in Cognitive

Radio Games

Jong-Shi Pang Industrial & Enterprise Systems Engineering, University of Illinois at Urbana-Champaign

We propose a game-theoretic approach to study a cognitive radio (CR) paradigm in signal processing composed of finitely many primary users (PUs) and secondary users (SUs), wherein each SU (player) competes against the others to maximize his own opportunistic throughput by choosing jointly the sensing duration, the detection thresholds, and the power allocation over a multichannel spectrum. The game contains constraints on the transmit power (and possibly spectral masks) as well as the maximum individual/aggregate (probabilistic) average interference tolerable by the PUs. To keep the optimization as decentralized as possible, global interference constraints are imposed via pricing. The individual players' optimization problems are nonconvex and there are price clearance conditions associated with the nonconvex global interference constraints to be satisfied by the Nash equilibria of the game. Using variational inequality theory, we establish the existence of an aggregated first-order solution of the game, which we call a Quasi-Nash Equilibrium (QNE), and relate such a QNE to a local Nash Equilibrium (LNE). We further establish the existence of a Nash equilibrium solution under a certain smallness restriction on the interference coefficients between the PUs and SUs. Lastly, we present provably convergent distributed algorithms for computing such equilibria along with their convergence. Numerical results show that the proposed game-theoretic approach of joint sensing and power allocation yields considerable performance improvement with respect to current centralized and decentralized designs of CR systems.

This work is joint with Gesualdo Scutari at the State University of New York at Buffalo. Notes:

9:45-10:45, July 12


Optimization of Cooperative Wireless Multicell

Networks

Wei Yu University of Toronto

The mitigation of intercell interference is a central issue for future generation wireless cellular networks where frequencies are reused aggressively and where hierarchical cellular structures may heavily overlap. In this talk, we examine the benefit of multicell cooperation for interference mitigation. The first part of the talk deals with coordination at the transmission strategy and resource allocation level. We show that a joint proportionally fair scheduling, spatial multiplexing, and power spectrum adaptation method that coordinates multiple base-stations can significantly improve the overall network throughput while maintaining user fairness. The second part of this talk deals with signal-level coordination in a network MIMO model where the base-stations are connected via high-capacity backhaul links. We characterize the tradeoff between downlink sum-rate gain and the backhaul capacity, and illustrate the benefit of dynamic cooperation link selection, user scheduling, precoding, and power spectrum optimization. Notes:

11:30-12:30, July 12


Physical-Layer Multicasting for MISO Downlink:

Going beyond Transmit Beamforming

Wing-Kin (Ken) Ma Department of Electronic Engineering, The Chinese University of Hong Kong

Physical-layer multicasting has recently received growing attention in standards such as LTE. It is a physical-layer or signaling technique for broadcasting common information to a specific group of users, given their channel state information at the transmitter; e.g., real-time mobile TV streaming to a group of paid subscribers. A well-known, significant, approach in this context is transmit beamforming via semidefinite relaxation (SDR), first advocated by Sidiropoulos, Davidson, and Luo. While multicast transmit beamforming is an effective approach, especially for not too many number of users, there are few works that go beyond transmit beamforming. This talk will present two new multicast strategies. The first strategy, called stochastic beamforming (SBF), employs a random-in-time philosophy to circumvent the rank-one approximation issue inherent in SDR-based beamforming. The second strategy is a combination of beamforming and the Alamouti space-time code; this results in a rank-two generalization of the previous SDR-based beamforming scheme. By theoretical analysis, we prove that these two strategies can work better than SDR-based beamforming—SBF can achieve a constant achievable rate gap of 0.8314 bits/s/Hz relative to the optimal multicast capacity, irrespective of the number of users, while the beamformed Alamouti scheme has a provably better worst-case SNR scaling than beamforming. These claims are corroborated by simulation results, wherein it is demonstrated that under a many-users setting, the two proposed strategies can substantially outperform beamforming in bit error probability performance Notes:

14:30-15:30, July 12


Algorithms on random lattices in wireless

communications

Joakim Jalden KTH Royal Institute of Technol

Many problems in wireless communications can be phrased in terms of point lattices, i.e., periodic arrangements of points in Euclidean space. An important example is the optimal decoding problem in space-time coded multiple antenna wireless communications which is closely connected to the classical closest vector problem (CVP) in computer science. The optimal decoding problem however also unfortunately shares the NP-hard nature of the CVP making it a challenging task, and the object of much research. Over the past decade, researchers in wireless communication have also increasingly relied on mathematical results for lattices to construct of computationally efficient approximations of the optimal decoder. However, unlike in many classical areas involving lattices, the lattices in wireless communication are randomly generated and this calls for new analysis tools.

This talk will illustrate how classical results on the theory of point lattices, optimization, and probability theory - in particular random matrix theory and the theory of large deviations, may be combined to analyze and design receivers in wireless communication. We will present and explain some existing results, and provide an overview of open problems and research directions in the area. Notes:

15:30-16:30, July 12


Interference Calculus -- A General Framework for

Interference Management and Network Utility

Optimization

Martin Schubert Technical University Berlin

The analysis and optimization of multi-user communication networks is complicated by the presence of mutual interference and limited resources. A typical example is a wireless system, where users can avoid or mitigate interference in a flexible way by dynamically adjusting the resources allocated to each user. In addition, adaptive techniques for interference mitigation (e.g. beamforming, MIMO) can cause the interference to depend on the underlying resources in a complicated nonlinear fashion. For the current development of dense, high-rate wireless systems, interference emerges as a key performance-limiting factor.

Interference Calculus is a framework for modeling and optimizing interference-coupled networks. At the core of this theory lies the concept of homogeneous monotone interference functions. This basic framework is general enough to be applicable to a wide range of scenarios while still providing enough structure for the analysis and optimization.

In this talk, we give an overview on the theory of interference functions and applications. We discuss how the framework generalizes known concepts and results from power control theory, and in particular the Perron-Frobenius theory of non-negative matrices. Then, it is shown how properties of interference functions (namely monotonicity, convexity, and log-convexity) can be exploited for the design of globally optimal algorithms. The connection between monotone interference functions and comprehensive utility sets is discussed. Finally, strategies for achieving fairness and efficiency are presented. Notes:

8:45-9:45, July 13


Multichannel Acoustic Signal Processing for Spatial

Sound Acquisition and Delivery

Jingdong Chen Northwestern Polytechincal University

Conferencing capability is an integral part of modern communication networks. It facilitates group collaborations efficiently in an economic way. A key technical challenge for a teleconferencing (or more generally a telecollaboration) system is the ability to acquire and deliver high-fidelity spatial sound so that it is possible for the remote listener to follow a panel of speakers and distinguish them by listening to the reproduced signals. To acquire and deliver spatial sound, it is necessary to use multiple microphones and multiple loudspeakers in both the transmission and receiving ends. The system in each end then becomes a multi-input-multi-output (MIMO) one that is more complicated and difficult to solve than a single-channel system. Consequently, many acoustic signal processing problems need to be readdressed such as noise reduction, echo cancellation/control, source separation, dereverberation, to name a few. In this talk, I will address the problem and challenges of MIMO acoustic signal processing, with emphasis being placed on the MIMO channel identification and noise reduction. I will elaborate, using examples, on how to extract the desired speech signals and mitigate the effect of noise and interference while preserving the spatial information of the desired sound sources. Notes:

9:45-10:45, July 13


New Advances on Radar Detection in Presence of

Non-Gaussian Clutter

Maria Greco University of Pisa

The modeling of the clutter echoes is a central issue for the design and performance evaluation of radar systems. Aim of this talk is to describe the state-of-the-art approaches to the modeling and understanding of non-Gaussian radar clutter echoes and their implications on performance prediction and signal processors design.

The first part of the talk will be dedicated to modern statistical and spectral models for high resolution sea and ground clutter and to the methods of experimental validation using recorded data sets.

Coherent radar target detection in non-Gaussian clutter is the subject of the second part of the talk. In high resolution radar systems the disturbance cannot be modelled as Gaussian distributed and the classical detectors suffer from high losses. Then, according to the adopted disturbance model, optimum and sub-optimum detectors are derived and their performance analyzed against a non-Gaussian background. Different interpretations of the various detectors are provided to highlight the relationships and the differences among them. In particular, it is shown how the Generalized Likelihood Ratio Test (GLRT) detector may be recast into the form of the generalized whitening matched filter (GWMF), which is the GLRT detector against Gaussian disturbance, compared to a data-dependent threshold. The proposed detectors are tested against both simulated data and measured high resolution sea clutter data to investigate the dependence of their performance on the various clutter and signal parameters. Notes:

11:30-12:30, July 13


Phase Unwrapping Problem

Ming Jiang and Wenchang Sun

School of Mathematical Sciences, Peking University Department of Mathematics, Nankai University

Phase unwrapping is a classical problem which arises from many branches of applied physics and engineering, such as optical interferometry, x-ray phase contrast imaging, MRI, and synthetic aperture radar, etc. In these applications, true phase values from an imaged object are “wrapped” in one principal range such as $(-\pi , pi]$. This results in not only discontinuities in the measured phase values, but also ambiguities of phase values. Phase unwrapping problem is to restore true phase values from the measured phase data which are wrapped and corrupted by noise. Phase unwrapping is essentially ill-posed due to the ambiguity resulted from the wrapping operator. In our previous work, we developed algorithms for 2-dimensional phase unwrapping problem. One is based on the integer optimization approach and one is based on the Neumann problem of the Poisson’s equation.

In this talk, we will discuss the sampling issues with the phase unwrapping problem. We have established a sampling formula when the phase gradient is band-limited and its gradient is less than $\pi$-times of the Nyquist frequency. When this sampling condition on the gradient fails, or when the phase is not band-limited, we have estimated reconstruction errors for band-limited, or exponentially or polynomially decayed phases, respectively, in terms of the sampling interval. Notes:

9:45-10:45, July 13


14:30-15:30, July 13


Quantitative SNR Studies for Joint Time-Frequency

Analysis with Applications in SAR/ISAR Imaging

Xiang-Gen Xia Dept of Electrical and Computer Engineering, University of Delaware

Joint time-frequency analysis (JTFA) has been studied extensively in the past decades and found many applications in signal and image processing, such as radar signal processing etc. A JTFA usually localizes a signal, such as a chirp, in the joint time-frequency (TF) plane, while it spreads noise over the TF plane. This implies that a JTFA usually increases the signal-to-noise ratio (SNR). In this talk, we quantitatively analyze the SNR increase rate in the joint TF plane over the SNR in the time domain or in the frequency domain. We then apply it to ISAR imaging of maneuvering targets. Notes:

15:30-16:30, July 13


Average Sampling and Signal Denoising

Wenchang Sun Nankai University

We introduce some recent advances in sampling theory, which include average, local, and generalized sampling. Based on the theory of average sampling, we present a new algorithm to reconstruct bandlimited signals from sampled values in the presence of zero mean, independent and identically distributed random noises. Notes:


Notes:

Map of Xidian University

contentsmeeting.xidian.edu.cn/.../20120610/20120610134501.pdf · and optimization from academia,...

Documents