Qing Ling - Research

Our research focuses on decentralized network optimization and its applications. This is a world of networks. Various networks around us are collecting data and processing data; examples include wireless sensor and actuator networks, vehicular ad hoc networks, autonomous robot networks, mobile device networks, distributed or decentralized computing networks, etc. Within these networks, nodes are subject to limited energy, communication, and computation resources. There are often no fusion centers to coordinate actions of the nodes. With the aid of optimization tools, we devote to design and analyze efficient network data processing algorithms that are faithful to the resource-limited and decentralized natures of the networks.

Development and Convergence Rate Analysis of Decentralized Optimization Algorithms. We develop decentralized optimization algorithms based on the classic optimization tools, such as first-order methods, second-order methods, splitting methods, etc. We are particularly interested in their convergence rates that directly influence communication and coordination costs in a network. We shed light on the connection between convergence rate and network topology and provide guidience to network design and parameter tuning.

Decentralized Dynamic and Online Optimization. When each node observes time-varying data or one piece of data at each time instance, the network needs to solve a decentralized dynamic or online optimization problem. We develop decentralized dynamic and online optimization algorithms and investigate their convergence in terms of network topology and data dynamics.

Privacy Preservation in Decentralized Optimization. One remarkable advantange of decentralized optimization is that nodes can hold their private data and avoid transmitting them to a fusion center. This scheme protects data privacy against a malicious or insecure fusion center. In our research, we analyze privacy-preservation against malicious nodes in decentralized optimization.

Sparse Optimization. By exploiting structures (sparsity, group-sparsity, low-rankness) of solutions, sparse optimization not only helps formulate otherwise ill-conditioned inverse problems but also enables development of efficient algorithms. We analyze identifiability of sparse optimization models, design customized algorithms, and discuss its applications in video processing and direction of arrival estimation.

Decentralized Sensing and Control with Wireless Sensor and Actuator Networks. Traditional sensing and control systems rely on fusion centers and central controllers to collect data and make decisions that is neither robust nor scalable in a network environment. We are developing a prototype wireless sensor and actuator network in an experimental greenhouse.

Environmental Monitoring and Self-localization of Wireless Sensor Networks. Environmental monitoring is one of the main application areas of wireless sensor networks, while self-localization is one of the main enabling technologies for wireless sensor networks. We develop energy-efficient decentralized environmental monitoring algorithms and error-tolerant self-localization algorithms, as well as implement them in practical wireless sensor networks.

Vehicular Ad Hoc Networks. In a vehicular ad hoc network, moving cars or buses are nodes that create a communication network. We focus on trustable information aggregation under the dynamic network topology. We have developed a hierarchical wireless monitoring system for pure electric buses that are running in the public transportation system in Hefei. We are now developing a decentralized vehicular ad hoc network.

EXTRA: An Exact First-order Algorithm for Decentralized Consensus Optimization
School of Sciences, Hangzhou Dianzi University, Hangzhou, China, 2017.6.14

Learn-and-Adapt Stochastic Dual Gradients for Network Resource Allocation
International Conference on Engineering and Computational Mathematics, Hong Kong, China, 2017.6.1

Decentralized Consensus Optimization with Asynchrony and Delays
School of Mathematical Sciences, University of Electronic Science and Technology of China, Chengdu, China, 2017.4.22

EXTRA: An Exact First-Order Algorithm for Decentralized Consensus Optimization
School of Data and Computer Science, Sun Yat-Sen University, Guangzhou, China, 2017.1.5

Decentralized Consensus Optimization
School of Computer Science and Technology, Guangdong University of Technology, Guangzhou, China, 2017.1.4

Collaborative Resource Allocation over a Hybrid Cloud Center and Edge Server Network
The 2016 Workshop on Frontiers of Optimization Theories and Methods, Linyi, China, 2016.5.7

Learning Deep L0 Encoders
The 2016 Mini-Symposium on Large-Scale Optimization for Data Analytics, Shanghai, China, 2016.3.20

Decentralized Network Optimization: Algorithms and Theories
Institute of Systems Science, Chinese Academy of Sciences, Beijing, China, 2015.12.14

Decentralized Network Optimization: Algorithms and Theories
Institute of Computational Mathematics, Chinese Academy of Sciences, Beijing, China, 2015.12.12

Learning Deep L0 Encoders
The 2015 Youth Symposium of Scientific and Engineering Computation, Beijing, China, 2015.12.11

EXTRA: An Exact First-order Algorithm for Decentralized Consensus Optimization
Department of Mathematics, Hong Kong University of Science and Technology, Hong Kong, China, 2015.11.10

Decentralized Network Optimization: Algorithms and Theories
The State Key Laboratory of Rail Traffic Control and Safety, Beijing Jiaotong University, Beijing, China, 2015.8.13

EXTRA: An Exact First-order Algorithm for Decentralized Consensus Optimization
Department of Control Science and Engineering, Zhejiang University, Hangzhou, China, 2015.4.10

Decentralized Dynamic Optimization through the Alternating Direction Method of Multipliers
The 2014 International Workshop on Signal Processing, Optimization, and Compressive Sensing, Changsha, China, 2014.12.22

EXTRA: An Exact First-order Algorithm for Decentralized Consensus Optimization
Institute of Computational Mathematics, Chinese Academy of Sciences, Beijing, China, 2014.12.1

Decentralized Dynamic Optimization through the Alternating Direction Method of Multipliers
Institute of Computational Mathematics, Chinese Academy of Sciences, Beijing, China, 2014.10.17

Decentralized Optimization for Multi-agent Networks
The 2014 Workshop on Optimization for Modern Computation, Beijing, China, 2014.9.2

Decentralized (Linearized) Alternating Direction Method of Multipliers
The 2014 SIAM Conference on Imaging Science, Hong Kong, China, 2014.5.14

Linearly-Convergent Decentralized Algorithms for Multi-agent Network Optimization
Digital Technology Center, University of Minnesota, Twin Cities, USA, 2014.1.23

Linearly-Convergent Decentralized Algorithms for Multi-agent Network Optimization
Department of Electrical and Computer Engineering, Iowa State University, Ames, USA, 2014.1.22

Linearly-Convergent Decentralized Algorithms for Multi-agent Network Optimization
Department of Electrical Engineering and Computer Science, Syracuse University, Syracuse, USA, 2013.10.16

Decentralized Jointly Sparse Optimization by Reweighted Lq Minimization
Department of Electrical and Computer Engineering, University of Houston, Houston, USA, 2012.9.5

Decentralized Low-rank Matrix Completion
The 2012 International Workshop on Signal Processing, Optimization, and Control, Hefei, China, 2012.7.2

Decentralized Jointly Sparse Optimization by Reweighted Lq Minimization
The 2012 National Conference on Mathematical Programming of China, Hangzhou, China, 2012.4.21

Decentralized Optimization of Networked Multi-agent Systems
Department of Mathematics, Hong Kong Baptist University, Hong Kong, China, 2011.7.26

Decentralized Optimization in Wireless Sensor Networks
Institute of Computational Mathematics, Chinese Academy of Sciences, Beijing, China, 2010.4.9