Article
Mathematics, Applied
Huan Ren, Ru-Ru Ma, Qiaohua Liu, Zheng-Jian Bai
Summary: In this paper, a randomized quaternion QLP decomposition algorithm is proposed for computing a low-rank approximation to a quaternion data matrix, based on quaternion normal distribution random sampling. The convergence results and upper bounds of the proposed algorithm outperform existing methods for real QLP decomposition, and the algorithm can track the singular values of the quaternion data matrix with high probability.
JOURNAL OF SCIENTIFIC COMPUTING
(2022)
Article
Biochemical Research Methods
Kuo Yang, Yuxia Yang, Shuyue Fan, Jianan Xia, Qiguang Zheng, Xin Dong, Jun Liu, Qiong Liu, Lei Lei, Yingying Zhang, Bing Li, Zhuye Gao, Runshun Zhang, Baoyan Liu, Zhong Wang, Xuezhong Zhou
Summary: Drug repositioning is an important method in drug development that identifies new indications of approved drugs through analysis of clinical and experimental data. This study proposes a drug repositioning framework called DRONet, which combines network embedding and ranking learning to utilize the effectiveness comparative relationships among drugs. Experimental results show that DRONet achieves higher prediction accuracy than existing methods and demonstrates potential for guiding clinical drug development.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Biochemical Research Methods
Xinxing Yang, Genke Yang, Jian Chu
Summary: Computational drug repositioning aims to discover new therapeutic diseases for marketed drugs, providing advantages over traditional drug development in terms of cost, development cycle, and controllability. The matrix factorization model is widely used due to its easy implementation and scalability, but lacks expressive ability in representing the association between drugs and diseases. To overcome this, a neural metric factorization model (NMFDR) is proposed, considering the latent factor vectors of drugs and diseases as points in a high-dimensional coordinate system and introducing a generalized euclidean distance to represent their association. By embedding multiple drug (disease) metrics information into the latent factor vectors, the similarity between drugs (diseases) can be reflected in the distance between them. Experimental analysis on real datasets demonstrates the effectiveness and superiority of the NMFDR model.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2023)
Article
Mathematics, Applied
James Demmel, Laura Grigori, Alexander Rusciano
Summary: We introduce a Generalized LU Factorization (GLU) for low-rank matrix approximation and compare it with past approaches. We provide complete bounds by combining established deterministic guarantees with sketching ensembles satisfying Johnson-Lindenstrauss properties. The subsampled randomized Hadamard transform (SRHT) ensemble shows particularly good performance. Moreover, the factorization unifies and generalizes many past algorithms, sometimes providing strictly better approximations, and helps explain the effect of sketching on the growth factor during Gaussian elimination.
SIAM JOURNAL ON MATRIX ANALYSIS AND APPLICATIONS
(2023)
Article
Mathematics, Applied
Maolin Che, Yimin Wei, Hong Yan
Summary: This paper focuses on developing randomized algorithms for computing low multilinear rank approximations of tensors using random projection and singular value decomposition. The probabilistic analysis of error bounds for the randomized algorithm is based on the theory of singular values of sub-Gaussian matrices. The effectiveness of the proposed algorithms is demonstrated through several numerical examples.
JOURNAL OF COMPUTATIONAL AND APPLIED MATHEMATICS
(2021)
Article
Materials Science, Multidisciplinary
Jason Kaye, Kun Chen, Olivier Parcollet
Summary: We present an efficient basis for imaginary time Green's functions based on a low-rank decomposition of the spectral Lehmann representation. The basis functions are simply a set of well-chosen exponentials, and the corresponding expansion can be considered as a discrete form of the Lehmann representation. The basis is determined by an upper bound on the product beta omega(max), the inverse temperature and energy cutoff, and a user-defined error tolerance. The number of basis functions scales logarithmically with the product beta omega(max) and the reciprocal of the error tolerance. The basis functions and interpolation nodes can be obtained rapidly using standard numerical linear algebra routines. The discrete Lehmann representation of the Green's function can be transformed to the Matsubara frequency domain or obtained directly by interpolation on a Matsubara frequency grid.
Article
Telecommunications
Ziping Wei, Haozhan Li, Hongfu Liu, Bin Li, Chenglin Zhao
Summary: This paper proposes a novel CSI feedback method with low complexity and high precision to accurately recover CSI. The method exploits the low-rank characteristic of a large channel matrix by approximating it with small sub-matrices.
IEEE COMMUNICATIONS LETTERS
(2021)
Article
Mathematics, Applied
Hong Zhu, Michael K. Ng, Guang-Jing Song
Summary: A new approximate method for solving nonnegative low-rank matrix approximation problem is developed in this study. It involves alternately projecting onto fixed-rank matrix manifold and nonnegative matrix manifold to ensure convergence, with numerical results demonstrating its performance.
JOURNAL OF SCIENTIFIC COMPUTING
(2021)
Article
Mathematics, Applied
David Persson, Daniel Kressner
Summary: This work presents a method called funNystrom for computing low-rank approximations of a matrix function f(A) for a large symmetric positive semidefinite matrix A. Compared to other randomized methods, funNystrom avoids matrix-vector products with f(A) and achieves comparable results to the best low-rank approximation with lower computational cost.
SIAM JOURNAL ON MATRIX ANALYSIS AND APPLICATIONS
(2023)
Article
Engineering, Electrical & Electronic
Maboud F. Kaloorazi, Jie Chen
Summary: A new algorithm called PbP-QLP is introduced in this paper for efficiently approximating low-rank matrices without using pivoting strategy, which allows it to leverage modern computer architectures better than competing randomized algorithms. The efficiency and effectiveness of PbP-QLP are demonstrated through various classes of synthetic and real-world data matrices.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2021)
Article
Computer Science, Information Systems
Maboud F. Kaloorazi, Kai Liu, Jie Chen, Rodrigo C. De Lamare, Susanto Rahardja
Summary: This article presents a new method called Randomized Unpivoted QLP (RU-QLP), which performs a partial QLP decomposition of matrices using random sampling and unpivoted QR decomposition. RU-QLP achieves fast and parallel computation while preserving the rank-revealing property of pivoted QLP.
Article
Engineering, Electrical & Electronic
Ziping Wei, Hongfu Liu, Bin Li, Chenglin Zhao
Summary: This paper proposes a novel computation and communication efficient scheme for acquiring channel state information at transmitter (CSIT) in a massive MIMO system. By jointly designing downlink CSI estimation and uplink feedback, the scheme reduces computation complexity and communication overhead while achieving high accuracy.
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
(2022)
Article
Engineering, Electrical & Electronic
Farnaz Sedighin, Andrzej Cichocki, Anh-Huy Phan
Summary: The paper introduces a new rank selection method for TR decomposition, which gradually increases TR rank sizes in each iteration and selects core tensors based on their sensitivity to approximation errors, leading to a significant reduction in storage costs while maintaining the desired approximation accuracy.
IEEE JOURNAL OF SELECTED TOPICS IN SIGNAL PROCESSING
(2021)
Article
Mathematics, Applied
Arvind K. Saibaba, Rachel Minster, Misha E. Kilmer
Summary: This paper introduces a method for multivariate function approximation using Chebyshev polynomials and tensor compression techniques, and provides detailed analysis and experiments. The results show that the proposed method has significant advantages in terms of computational and storage efficiency.
ADVANCES IN COMPUTATIONAL MATHEMATICS
(2022)
Article
Engineering, Electrical & Electronic
Hamid Fathi, Emad Rangriz, Vahid Pourahmadi
Summary: This study proposes two algorithms to recover coherent and incoherent low-rank matrices, where a limited sampling budget is used to sample the most informative elements based on uncertainty information metric. Simulation results demonstrate the superior performance of the proposed algorithms compared to traditional methods.
IEEE SIGNAL PROCESSING LETTERS
(2021)
Article
Computer Science, Information Systems
Weihe Li, Jiawei Huang, Shiqi Wang, Chuliang Wu, Sen Liu, Jianxin Wang
Summary: Video traffic has been growing exponentially due to the increasing popularity of mobile devices and network improvements. Most commercial players use adaptive bitrate algorithms to choose bitrates based on network capacity and buffer occupancy. However, current algorithms prioritize average bitrate over perceptual quality, resulting in a degraded Quality of Experience (QoE). To tackle this issue, we propose DAVS (Dynamic-chunk quality Aware Video Streaming), an adaptive bitrate algorithm that employs apprenticeship learning to select higher quality for dynamic chunks without sacrificing static chunk quality excessively. Additionally, DAVS takes into account user viewing preferences to adapt to QoE diversity. Experimental results demonstrate that DAVS improves the quality of dynamic chunks and significantly enhances QoE compared to other representative algorithms.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Chemistry, Medicinal
Teng-Zhi Long, Shao-Hua Shi, Shao Liu, Ai-Ping Lu, Zhao-Qian Liu, Min Li, Ting-Jun Hou, Dong-Sheng Cao
Summary: This study constructed a high-quality dataset and established a series of classification models using machine learning algorithms to predict hematotoxicity. The best model based on Attentive FP showed excellent performance on both the validation and test sets. Additionally, the study utilized SHAP and atom heatmap methods to identify important features and structural fragments related to hematotoxicity, and employed MMPA and representative substructure derivation technique to further investigate the transformation principles and distinctive structural features of hematotoxic chemicals.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2023)
Article
Neurosciences
Bingbin Liu, Yuxuan Qian, Jianxin Wang
Summary: This study developed a multi-rodent tracking system (EDDSN-MRT) to investigate the relationship between neural systems and social behaviors. The EDDSN-MRT system can track the locomotion and social behavior of multiple mice simultaneously and provides better tracking performance compared to other methods.
Article
Biology
Chuqi Lei, Zhangli Lu, Meng Wang, Min Li
Summary: A study proposes an ensemble learning model called StackCPA, which accurately predicts compound-protein binding affinity by integrating multi-scale features of protein pocket and compound through a transfer learning strategy. The experiment results show that StackCPA outperforms other state-of-the-art deep learning models on three benchmark datasets. The protein pocket provides sufficient information for affinity prediction, and its multi-scale features further improve the prediction performance.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Biochemical Research Methods
Min Li, Wenbo Shi, Fuhao Zhang, Min Zeng, Yaohang Li
Summary: Understanding protein functions is crucial for various biological problems, but there is a large gap between the increase in protein sequences and the annotations of protein functions. To address this issue, a new deep learning model, DeepPFP-CO, uses Graph Convolutional Network (GCN) to explore and capture the co-occurrence of GO terms, improving protein function prediction performance.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2023)
Article
Computer Science, Information Systems
Zhaoyi Li, Jiawei Huang, Jinbin Hu, Weihe Li, Tao Zhang, Jingling Liu, Jianxin Wang, Tian He
Summary: This article presents a new receiver-driven congestion control design called REN, which addresses the challenges brought by network dynamic and achieves ultra-low latency. REN utilizes under- and over-utilization notifications from switches to handle dynamic traffic, mitigates burstiness and conservativeness, and effectively reduces average flow completion time (AFCT) by up to 68%.
IEEE TRANSACTIONS ON CLOUD COMPUTING
(2023)
Article
Biochemical Research Methods
Linyuan Guo, Tian Qiu, Jianxin Wang
Summary: Protein-ligand interactions are crucial for cellular activities and drug discovery. The complexity and cost of experimental methods have led to a demand for computational approaches for deciphering protein-ligand interaction patterns. This study introduces a deep learning-based scoring function called ViTScore, which accurately identifies near-native poses from a set of poses. ViTScore shows promise as a tool for protein-ligand docking, accurately identifying potential drug targets and aiding in the design of new drugs with improved efficacy and safety.
IEEE TRANSACTIONS ON NANOBIOSCIENCE
(2023)
Article
Computer Science, Theory & Methods
Jingui Huang, Jie Chen, Yunlong Liu, Guang Xiao, Jianxin Wang
Summary: In this paper, we study the fixed-order book drawing problem and develop algorithms for it from the perspective of parameterized complexity. By limiting the number of crossings per edge and other parameters of the input graph, we obtain specific results.
INTERNATIONAL JOURNAL OF FOUNDATIONS OF COMPUTER SCIENCE
(2023)
Article
Biochemical Research Methods
Siqi Chen, Ruiqing Zheng, Luyi Tian, Fang-Xiang Wu, Min Li
Summary: scRNA-seq data often have many zeros, and these dropout events hinder downstream data analysis. We propose BayesImpute to infer and impute dropout values in scRNA-seq data. By utilizing the expression rate and coefficient of variation of genes, BayesImpute identifies likely dropouts and imputes missing values using posterior distributions. Simulated and real experiments show that BayesImpute effectively identifies dropout events, reduces false positive signals, and preserves biological information. Furthermore, BayesImpute improves cell subpopulation analysis, differential expression detection, and is scalable and fast with low memory usage compared to other statistical-based imputation methods.
Article
Computer Science, Information Systems
Wenkang Wang, Xiangmao Meng, Ju Xiang, Yunyan Shuai, Hayat Dino Bedru, Min Li
Summary: Protein complexes are essential in living cells and detecting them is crucial for understanding protein functions and treating complex diseases. This study proposes a novel method, CACO, to detect human protein complexes by integrating functional information from other species via protein ortholog relations. The method outperforms other state-of-the-art methods in terms of F-measure and Composite Score, demonstrating the effectiveness of integrating ortholog information and the proposed core-attachment algorithm in detecting protein complexes.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Computer Science, Information Systems
Yixiong Liang, Shuo Feng, Qing Liu, Hulin Kuang, Jianfeng Liu, Liyan Liao, Yun Du, Jianxin Wang
Summary: To improve the performance of cervical abnormal cell detection, we propose a method that utilizes contextual relationships. By exploring the relationships between cells and cell-to-global images, the features of each region of interest are enhanced. Our experiments on a large dataset validate the effectiveness of the proposed method by achieving better average precision (AP) than baseline methods.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Computer Science, Information Systems
Hulin Kuang, Yahui Wang, Yixiong Liang, Jin Liu, Jianxin Wang
Summary: This study proposes a new body and edge aware network for medical image segmentation. It introduces various modules and applies deep supervision to effectively extract body and edge features, resulting in improved segmentation performance.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Computer Science, Information Systems
Jinbin Hu, Jiawei Huang, Zhaoyi Li, Jianxin Wang, Tian He
Summary: Existing congestion control protocols in datacenter networks struggle to simultaneously achieve ultra-low latency and high link utilization across all types of workloads. We propose AMRT, an Anti-ECN Marking Receiver-driven Transport protocol, which achieves near-zero queueing delay and full link utilization by increasing sending rate when under-utilization is detected. Experimental results demonstrate that AMRT reduces the average flow completion time (AFCT) by up to 42% and improves link utilization by up to 38% compared to state-of-the-art receiver-driven transmission schemes.
IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT
(2023)
Article
Biology
Haochen Zhao, Peng Ni, Qichang Zhao, Xiao Liang, Di Ai, Shannon Erhardt, Jun Wang, Yaohang Li, Jianxin Wang
Summary: Adverse drug reactions have a direct impact on human health. Computational methods, such as the deep learning framework GCAP, offer promising alternatives for predicting the seriousness of clinical outcomes resulting from adverse reactions to drugs. GCAP can effectively predict whether adverse reactions cause serious clinical outcomes and infer the corresponding classes of seriousness.
COMMUNICATIONS BIOLOGY
(2023)
Article
Biochemical Research Methods
Yiwei Liu, Changhuo Yang, Hong-Dong Li, Jianxin Wang
Summary: The article presents a feature selection-based approach named IsoFrog for predicting isoform functions. It uses a reversible jump Markov Chain Monte Carlo (RJMCMC) feature selection framework and a sequential feature selection procedure to improve the accuracy of the model.