Article
Mathematics, Applied
Ruoning Chen, Liping Zhang, Zhenyu Ming
Summary: Internet traffic network data recovery via link flow measurement is a challenging problem in wireless communication network due to the rapid growth of problem scale. In this study, a novel parallel low-rank matrix optimization model is constructed to accurately and quickly recover internet traffic network data. The proposed method outperforms the state-of-the-art methods in terms of recovery accuracy and computational cost, as demonstrated by numerical experiments on various datasets.
JOURNAL OF COMPUTATIONAL AND APPLIED MATHEMATICS
(2023)
Article
Engineering, Civil
Pallaviram Sure, Chithra Priya Srinivasan, C. Narendra Babu
Summary: The paper discusses the importance of road traffic sensing in Intelligent Transportation Systems (ITS) and the use of Low Rank Matrix Completion (LR-MC) approaches for accurate matrix reconstruction. Two new LR-MC approaches, AL-SRMF and C-LR, are developed and experimentally validated for traffic matrices of Californian road network. The results show that these new methods outperform existing approaches with significantly better performance.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Review
Computer Science, Artificial Intelligence
Zhanxuan Hu, Feiping Nie, Rong Wang, Xuelong Li
Summary: This paper provides a comprehensive survey on Low Rank Regularization (LRR), focusing on recent advances in rank-norm relaxation and model optimization. It emphasizes the superiority of non-convex relaxations over convex relaxations in improving the performance of existing LRR models.
Article
Engineering, Electrical & Electronic
Tong Wu
Summary: Tensor low-rank representation (TLRR) has been widely studied for capturing the intrinsic low-rank structure of tensor data. However, TLRR suffers from high computational complexity for large-scale data. In this paper, an online low-rank tensor subspace clustering algorithm (OLRTSC) is proposed to address this issue by leveraging the stochastic optimization technique. OLRTSC is online, can handle dynamic data, avoids computing t-SVD, and reduces storage cost significantly.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2023)
Article
Mathematics
Jun Sun, Lingchen Kong, Biao Qu
Summary: In recent years, the scientific community has become increasingly interested in re-identification of people, which remains a challenging problem due to low-quality images, occlusion between objects, and significant variations in lighting, viewpoint, and posture. We propose a dictionary learning method to reduce ambiguity in pedestrian visual characteristics by separating them into shared and specific parts. Experimental results demonstrate the effectiveness of our approach.
Article
Computer Science, Information Systems
Jianlou Xu, Yuying Guo, Wanqing Shang, Shaopei You
Summary: In this paper, a new image decomposition model based on deep learning is proposed. The deep image prior is used to describe the cartoon and the low-rank norm is used to describe the texture. Adaptive regularization parameters are employed to preserve the edge features. This is the first image decomposition model based on deep learning, and its validity is verified through numerical experiments.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Engineering, Multidisciplinary
Zi-Yue Zhu, Ting-Zhu Huang, Jie Huang
Summary: This paper proposes a new way to describe the low-rank prior by constructing a projection subspace and analyzing the abundance maps within it. Two algorithms are proposed based on different sparse structures, and experiments show their superiority compared to classical sparse unmixing algorithms.
APPLIED MATHEMATICAL MODELLING
(2024)
Article
Ergonomics
Ahmed Al-Kaisy, Kazi Tahsin Huda
Summary: This study investigates the application of the Empirical Bayes (EB) method and the Highway Safety Manual (HSM) predictive methodology for network screening on low-volume roads in Oregon. The findings reveal that the predicted number of crashes plays a major role in estimating the expected number of crashes on low-volume roads, and there is a significant discrepancy between the observed and predicted number of crashes using the HSM procedures. However, the expected number of crashes calculated using the HSM EB method is reasonably close to the observed number of crashes for the study sample.
JOURNAL OF SAFETY RESEARCH
(2022)
Article
Transportation Science & Technology
Tong Nie, Guoyang Qin, Yunpeng Wang, Jian Sun
Summary: Traffic speed is crucial for transportation applications, but current detection methods suffer from incomplete and noisy data. This study proposes a Laplacian enhanced low-rank tensor completion (LETC) framework to address this issue by incorporating multiple forms of correlations. Experimental results demonstrate that LETC achieves state-of-the-art performance and saves computing time compared to baseline methods.
TRANSPORTATION RESEARCH PART C-EMERGING TECHNOLOGIES
(2023)
Article
Computer Science, Artificial Intelligence
Wenjin Qin, Hailin Wang, Feng Zhang, Jianjun Wang, Xin Luo, Tingwen Huang
Summary: This paper proposes a framework for recovering order-d tensors, which achieves exact completion for any order-d low t-SVD rank tensors with missing values with an overwhelming probability. Experimental results demonstrate that the proposed method achieves highly competitive performance in terms of both qualitative and quantitative metrics.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Computer Science, Software Engineering
Milan Korda, Monique Laurent, Victor Magron, Andries Steenkamp
Summary: We introduce a new type of sparsity, called ideal-sparsity, for the generalized moment problem (GMP). By optimizing over a measure supported on the variety of an ideal generated by quadratic bilinear monomials, we obtain an equivalent sparse reformulation of GMP. The resulting hierarchies of moment-based relaxations for ideal-sparse reformulation provide tighter bounds and are computationally faster than the dense analogs.
MATHEMATICAL PROGRAMMING
(2023)
Article
Computer Science, Interdisciplinary Applications
Quan Chen, Huajun She, Yiping P. Du
Summary: A novel tensor dictionary learning algorithm, TDLLS, has been proposed to reconstruct myelin water content in the brain from undersampled T2* weighted images, improving the performance of tensor-based recovery. By incorporating low-rank constraints on the dictionaries and sparse constraints on the core coefficient tensors, the algorithm explores local and nonlocal similarity, and global temporal redundancy in the complex relaxation signals. Parallel imaging is applied for further acceleration, resulting in high-quality myelin water fraction maps obtained within 1 minute at an undersampling rate of 6.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2021)
Article
Physics, Multidisciplinary
Yoonhyuk Rah, Youngjae Jeong, Sangyoon Han, Kyoungsik Yu
Summary: Finding a reliable Ising machine for solving nondeterministic polynomial-class problems has attracted great attention. This research proposes an optomechanical coherent Ising machine with extremely low power consumption. By utilizing a new enhanced symmetry breaking mechanism and highly nonlinear mechanical Kerr effect, the power threshold is significantly reduced. This optomechanical spin model opens the possibility of chip-scale integration for large-size Ising machine implementations.
PHYSICAL REVIEW LETTERS
(2023)
Article
Operations Research & Management Science
Bin Gao, P-A Absil
Summary: In this paper, a Riemannian rank-adaptive method is proposed to address the low-rank matrix completion problem on a set of bounded-rank matrices. Numerical experiments demonstrate its superior performance compared to state-of-the-art algorithms, and show that each aspect of this rank-adaptive framework can be separately incorporated into existing algorithms for performance improvement.
COMPUTATIONAL OPTIMIZATION AND APPLICATIONS
(2022)
Article
Automation & Control Systems
Isaac Wilfried Sanou, Roland Redon, Xavier Luciani, Stephane Mounier
Summary: This article describes three online NN-CPD algorithms based on sparse dictionary learning for tracking chemical components in water. The methods take into account unknown factors and the variation of tensor rank, and utilize previous information to decompose upcoming new tensors. In addition, real-time acquisition of fluorescence data in a semi-controlled environment is also achieved in this study.
CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Bernardo Marenco, Paola Bermolen, Marcelo Fiori, Federico Larroca, Gonzalo Mateos
Summary: In this paper, we address the problem of online monitoring and detection of changes in the underlying data distribution using a graph representation learning method based on the RDPG model. We propose an efficient online change-point detection algorithm that quantifies the discrepancy between the streaming graph observations and the nominal RDPG. We provide insights on the algorithm's detection resolution and delay and offer an open-source implementation for weighted and directed graphs.
IEEE TRANSACTIONS ON SIGNAL AND INFORMATION PROCESSING OVER NETWORKS
(2022)
Article
Computer Science, Artificial Intelligence
Jun Sun, Tianyi Chen, Georgios B. Giannakis, Qinmin Yang, Zaiyue Yang
Summary: This paper proposes an adaptive communication method for the federated learning problem, which saves communication costs by quantizing gradients and skipping less informative communications. Extensive experiments validate the effectiveness of this method.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Vassilis N. Ioannidis, Siheng Chen, Georgios B. Giannakis
Summary: This study introduces an efficient pruned GST approach to address the complexity limitation of traditional GSTs. The pruned GSTs retain informative scattering features while bypassing exponential complexity, and achieve comparable performance to state-of-the-art GCNs.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Engineering, Electrical & Electronic
Abolfazl Hashemi, Rasoul Shafipour, Haris Vikalo, Gonzalo Mateos
Summary: This paper studies the problem of sampling and reconstructing spectrally sparse graph signals. It proposes efficient methods for different scenarios and analyzes the reconstruction error.
Article
Engineering, Electrical & Electronic
Manish K. Singh, Vassilis Kekatos, Georgios B. Giannakis
Summary: Recent works propose the use of deep neural networks to predict optimal power flow solutions in power systems applications. This paper introduces a sensitivity-informed DNN and demonstrates its effectiveness and constraint satisfaction capabilities in optimization problems.
IEEE TRANSACTIONS ON POWER SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Panagiotis A. Traganitis, Georgios B. Giannakis
Summary: This study explores the method of unsupervised ensemble classification and introduces novel algorithms for handling data dependencies in both sequential and networked data. Evaluation on synthetic and real datasets shows that knowledge of data dependencies in the meta-learner has a positive impact on unsupervised ensemble classification task.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2022)
Article
Automation & Control Systems
Manish K. Singh, Sairaj Dhople, Florian Doerfler, Georgios B. Giannakis
Summary: This letter proposes a Kron reduction method that eliminates nodes with zero current injections from electrical networks and applies it to the generalized setting of RL networks. Empirical tests on a Delta network are conducted to validate the analytical results.
IEEE CONTROL SYSTEMS LETTERS
(2022)
Proceedings Paper
Acoustics
Alireza Sadeghi, Georgios B. Giannakis
Summary: The use of federated learning framework allows for the distributed training of models on massive datasets without the need to upload data to a central server. However, maintaining performance and robustness in the face of varying data distributions across workers is a challenge. This study proposes a distributionally robust optimization framework and develops a primal-dual algorithm to ensure the trained model's robustness against adversarial attacks and distributional uncertainties.
2022 30TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO 2022)
(2022)
Proceedings Paper
Acoustics
Chang Ye, Gonzalo Mateos
Summary: In this study, a deep learning solution is proposed for localizing sources of network diffusion. By leveraging graph signal processing and the ADMM method, a diffusion filter and source locations can be estimated. The trained neural network model, SLoG-Net, is interpretable, parameter efficient, and offers controllable complexity, achieving comparable performance and significant speedups compared to traditional methods.
2022 30TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO 2022)
(2022)
Proceedings Paper
Acoustics
Yang Li, Gonzalo Mateos
Summary: Advances in graph signal processing contribute to the integration of brain structure and function in network neuroscience. This study proposes a supervised graph representation learning framework to model the relationship between brain structural connectivity and functional connectivity. The proposed framework effectively learns embeddings that preserve the similarity between brain networks, and demonstrates superior discriminative power in subject classification and visualization tasks.
2022 30TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO 2022)
(2022)
Proceedings Paper
Acoustics
Marcelo Fiori, Bernardo Marenco, Federico Larroca, Paola Bermolen, Gonzalo Mateos
Summary: The paper introduces the Random Dot Product Graph (RDPG) as a generative graph model for relational data and discusses the embedding problem of estimating latent positions from observed graphs. By utilizing recent advances in non-convex optimization, the paper proposes a first-order gradient descent method to solve the problem more effectively and demonstrates the effectiveness of the graph representation learning framework.
2022 30TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO 2022)
(2022)
Proceedings Paper
Automation & Control Systems
Manish K. Singh, D. Venkatramanan, Sairaj Dhople, Benjamin Kroposki, Georgios B. Giannakis
Summary: This paper outlines the energy-management problems for inverter-based power networks from the perspective of optimal control and (non)linear optimization. It categorizes the problems based on timescales and organizes them methodologically according to complexity. The paper establishes dynamic models, uncovers common assumptions, and postulates open challenges in this field.
Article
Engineering, Electrical & Electronic
Yang Li, Gonzalo Mateos, Zhengwu Zhang
Summary: Recent neuroimaging advances and algorithmic innovations enable the integration of brain structure and function, leading to the discovery of brain's organizing principles. This study develops a supervised graph representation learning framework to model the relationship between brain structural connectivity and functional connectivity, improving the accuracy of classifying alcohol consumption behavior.
IEEE TRANSACTIONS ON SIGNAL AND INFORMATION PROCESSING OVER NETWORKS
(2022)
Article
Computer Science, Theory & Methods
Yang Li, Gonzalo Mateos
Summary: This paper studies the macroscopic evolution of the football society from a complex network perspective. By analyzing football game records and constructing a football network, the dynamic features and community structures of the network are revealed. Furthermore, spatio-temporal analysis unveils the temporal states representing distinct development stages in football history.
APPLIED NETWORK SCIENCE
(2022)
Article
Engineering, Electrical & Electronic
Konstantinos D. Polyzos, Qin Lu, Georgios B. Giannakis
Summary: Semi-supervised learning (SSL) over graphs has gained popularity in network science applications. This work introduces a Bayesian SSL approach using Gaussian processes (GPs) to provide uncertainty quantification. An incremental learning mode is considered, and an ensemble of GP experts is utilized for prediction and weight updating. The random feature-based kernel approximation method is employed to ensure scalability and privacy preservation.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2022)
