Article
Computer Science, Artificial Intelligence
Qi Wang, Xiang He, Xu Jiang, Xuelong Li
Summary: Data clustering has attracted much attention, with various effective algorithms developed to handle the task. Non-negative matrix factorization (NMF) is considered powerful, but it has limitations in terms of sensitivity to noise and outliers. Existing graph-based NMF methods highly depend on the initial similarity graph and perform graph construction and matrix factorization separately, leading to suboptimal graph structures.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Genetics & Heredity
Xibo Sun, Leiming Cheng, Jinyang Liu, Cuinan Xie, Jiasheng Yang, Fu Li
Summary: This study developed a new method (LPI-WGRMF) for predicting lncRNA-protein interactions, demonstrating high-accuracy performance and providing suggestions for experimental validation of potential interactions.
FRONTIERS IN GENETICS
(2021)
Article
Biochemical Research Methods
Feng Zhou, Meng-Meng Yin, Cui-Na Jiao, Zhen Cui, Jing-Xiu Zhao, Jin-Xing Liu
Summary: This paper introduces an efficient computational method BGCMF, which utilizes collaborative matrix factorization to predict the associations between miRNAs and diseases. The BGCMF achieves desirable results with an AUC value up to 0.9514 in five-fold cross-validation experiments.
BMC BIOINFORMATICS
(2021)
Article
Mathematics, Interdisciplinary Applications
Laishui Lv, Dalal Bardou, Peng Hu, Yanqiu Liu, Gaohang Yu
Summary: This paper proposes a novel graph regularized nonnegative matrix factorization algorithm for temporal link prediction, which considers both the local and global information of the temporal network. Experimental results show that the proposed algorithm outperforms existing methods.
CHAOS SOLITONS & FRACTALS
(2022)
Article
Multidisciplinary Sciences
Siqi Peng, Akihiro Yamamoto, Kimihito Ito
Summary: We propose a new method for bipartite link prediction using matrix factorization with negative sample selection. This method outperforms previous unsupervised prediction methods and the raw MF-based method in two hypothetical application scenarios. The technique of negative sample selection helps in selecting reliable negative training samples in advance of the matrix factorization process, overcoming the unavailability of ground truth for absent links.
Article
Computer Science, Artificial Intelligence
Benhui Zhang, Xiaoke Ma
Summary: Multi-view clustering aims to assign objects into clusters with multiple heterogeneous features. However, current methods fail to capture the latent information of heterogeneous features, leading to suboptimal clustering performance. To address this issue, a novel algorithm is proposed to jointly learn the common representation and similarity structure in the embedding space, improving the performance of multi-view clustering.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Ming Gao, Xiangnan He, Leihui Chen, Tingting Liu, Jinglin Zhang, Aoying Zhou
Summary: Recent years have seen a surge of interest in network representation learning, with most research focusing on homogeneous or heterogeneous networks. However, there has been relatively little research on NRL for bipartite networks. This work introduces BiNE, a new solution that takes into account the unique properties of bipartite networks, such as the long-tail distribution of vertex degrees and implicit connectivity relations between vertices of the same type.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2022)
Article
Computer Science, Information Systems
Jin-Xing Liu, Zhen Cui, Ying-Lian Gao, Xiang-Zhen Kong
Summary: In recent years, the association between long non-coding RNAs (lncRNAs) and human diseases has attracted attention. However, predicting novel lncRNA-disease associations requires improvement in accuracy and effectiveness of methods.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2021)
Article
Computer Science, Artificial Intelligence
Ping Deng, Tianrui Li, Hongjun Wang, Shi-Jinn Horng, Zeng Yu, Xiaomin Wang
Summary: The objective of co-clustering is to identify similarity blocks between the sample set and feature set simultaneously. The nonnegative matrix tri-factorization algorithm is an important tool for co-clustering. To address the impact of noise, a tri-regularized NMTF model was proposed to optimize the performance and generalization ability of the model effectively.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Mathematics
Guowei Yang, Lin Zhang, Minghua Wan
Summary: This study proposed an exponential graph regularization non-negative low-rank factorization algorithm to improve the performance and robustness of NMF. By applying low-rank constraint, non-negative factorization, and graph embedding with matrix exponentiation, it aims to learn undisturbed latent data representations while preserving the local structure of known samples.
Article
Automation & Control Systems
Kexin Zhang, Xuezhuan Zhao, Siyuan Peng
Summary: A novel multiple graph regularized semi-supervised NMF method, MSNMF, is proposed in this paper, which combines limited supervised information in the form of pairwise constraints with multiple graph regularization to capture discriminative data representation. Experimental results on eight practical image datasets demonstrate that MSNMF can achieve better clustering results than several related NMF methods.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2021)
Article
Computer Science, Artificial Intelligence
Zhihao Huang, Zengfa Dou, Xiaoke Ma
Summary: The study proposes a nonnegative matrix factorization algorithm for multi-layer networks reduction, which improves the feature representation of network layers by jointly decomposing the topological structure and graph representation. Experimental results demonstrate that it is more accurate and robust than existing methods.
APPLIED SOFT COMPUTING
(2021)
Article
Chemistry, Medicinal
Iori Azuma, Tadahaya Mizuno, Hiroyuki Kusuhara
Summary: This study proposed a new method called neighborhood regularized bidirectional matrix factorization (NRBdMF) to predict drug effects by incorporating bidirectionality. The NRBdMF model achieved high accuracy and interpretability in predicting both side effects and therapeutic effects.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2023)
Article
Computer Science, Information Systems
Yuli Jiang, Huaijia Lin, Ye Li, Yu Rong, Hong Cheng, Xin Huang
Summary: This paper enhances the expressive power of GCN models by utilizing node features and building a node-feature bipartite graph. It exploits the bipartite graph convolutional network to model node-feature relations and achieves more accurate predictions for each node.
INFORMATION SCIENCES
(2023)
Article
Computer Science, Information Systems
Stuti Jain, Emilie Chouzenoux, Kriti Kumar, Angshul Majumdar
Summary: Simultaneous co-administration of multiple drugs may lead to adverse drug reactions. Identifying drug-drug interactions (DDIs) is crucial for drug development and repurposing of old drugs. This paper introduces a novel Graph Regularized Probabilistic Matrix Factorization (GRPMF) method that incorporates expert knowledge through graph-based regularization within a matrix factorization framework. An efficient optimization algorithm is proposed to solve the resulting non-convex problem. Evaluation using the DrugBank dataset demonstrates the superior performance of GRPMF compared to other techniques.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Computer Science, Information Systems
Zhongzhou Liu, Yuan Fang, Min Wu
Summary: Recommendation systems commonly overlook the popularity bias issue, which can impact the fairness of recommendations. To address this, the paper proposes a fairness-centric model, FAiR, that adaptively mitigates popularity bias for users and items. The model includes explicit fairness discriminators at the local level and an implicit discriminator at the global level, tailored to each individual user or item. Experimental results show that the model significantly outperforms state-of-the-art baselines in fairness metrics, while remaining competitive in effectiveness.
ACM TRANSACTIONS ON INFORMATION SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Yuecong Xu, Jianfei Yang, Haozhi Cao, Keyu Wu, Min Wu, Zhengguo Li, Zhenghua Chen
Summary: Multi-Source Domain Adaptation (MSDA) is a more practical scenario that relaxes the assumption in conventional Unsupervised Domain Adaptation (UDA). The existence of different domain shifts makes MSDA more challenging, especially in video domain where spatial-temporal features can cause negative transfer. To address this problem, TAMAN is proposed to dynamically align spatial and temporal feature moments for effective feature transfer.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Rajdeep Dutta, Harikumar Kandath, Senthilnath Jayavelu, Xiaoli Li, Suresh Sundaram, Daniel Pack
Summary: In this paper, a decentralized learning algorithm is proposed to restore communication connectivity in multi-agent formation control. The proposed scheme enables each mobile agent to raise the team connectivity by learning while connected to neighbors. When inter-agent communication is lost, a trained neural network generates control actions to restore connectivity. The approach leverages an adaptive control formalism and simulation results show its effectiveness even in the presence of velocity disturbances.
Article
Computer Science, Information Systems
Zeng Zeng, Ziyuan Zhao, Kaixin Xu, Yangfan Li, Cen Chen, Xiaofeng Zou, Yulan Wang, Wei Wei, Pierce K. H. Chow, Xiaoli Li
Summary: In this paper, a new correlation induced clustering method (CoIn) is proposed to address the problem of high-dimensional bioinformatics data clustering. The method captures complex correlations among high-dimensional data and guarantees correlation consistency within each cluster. The evaluation on a high-dimensional mass spectrometry dataset of liver cancer tumor demonstrates that the proposed method produces more explainable and understandable results for clinical analysis, showing the potential application for knowledge discovery in high-dimensional bioinformatics data.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Emadeldeen Eldele, Mohamed Ragab, Zhenghua Chen, Min Wu, Chee-Keong Kwoh, Xiaoli Li, Cuntai Guan
Summary: Sleep staging is crucial for diagnosing and treating sleep disorders. Current data-driven deep learning models for automatic sleep staging have limitations when dealing with real-world scenarios. To overcome these limitations, this study proposes a novel adversarial learning framework called ADAST, which addresses the domain shift problem in the unlabeled target domain. The proposed framework outperforms state-of-the-art methods in six cross-domain scenarios.
IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE
(2023)
Review
Biochemical Research Methods
Xingzhong Zhao, Anyi Yang, Zi-Chao Zhang, Yucheng T. Yang, Xing-Ming Zhao
Summary: Brain imaging genomics is an interdisciplinary field that integrates multimodal medical imaging features and multi-omics data, aiming to bridge the gap between macroscopic brain phenotypes and their molecular characteristics. Recent advances in large-scale imaging and multi-omics datasets have allowed the identification of common genetic variants associated with brain structure and function. Integrative analysis with functional multi-omics data has revealed critical genes, genomic regions, and cell types associated with brain phenotypes. This review highlights the importance of functional genomic datasets in understanding the biological functions of identified genes and cell types, and discusses the challenges and future directions in this field.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Biology
Youlin Zhan, Jiahan Liu, Min Wu, Chris Soon Heng Tan, Xiaoli Li, Le Ou-Yang
Summary: Detecting protein complexes is crucial for studying cellular organizations and functions. Existing computational methods for identifying protein complexes from protein-protein interaction (PPI) networks often ignore the signs of PPIs and do not consider joint clustering of multiple PPI networks. In this study, we propose a novel partially shared signed network clustering (PS-SNC) model that takes into account the signs of PPIs and can identify protein complexes from multiple state-specific signed PPI networks jointly.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Computer Science, Artificial Intelligence
Zhenghua Chen, Min Wu, Alvin Chan, Xiaoli Li, Yew-Soon Ong
Summary: Artificial Intelligence (AI) is a rapidly growing field that promises significant benefits for consumers and businesses, but its development has come at a cost to the environment and raised concerns about its societal impacts. This review explores machine learning approaches to address the sustainability problem of AI and proposes research challenges and directions for the next generation of sustainable AI techniques.
IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE
(2023)
Article
Engineering, Electrical & Electronic
Keyu Wu, Min Wu, Zhenghua Chen, Ruibing Jin, Wei Cui, Zhiguang Cao, Xiaoli Li
Summary: Domain adaptation transfers knowledge from label-rich source domains to label-scarce target domains for generalized learning in new environments. Partial domain adaptation (PDA) extends this concept by considering scenarios where the target label space is a subset of the source label space. This paper proposes a Reinforced Adaptation Network (RAN) that combines deep reinforcement learning with domain adaptation techniques to address the challenging PDA problem. Experimental results show that RAN significantly outperforms existing state-of-the-art methods on three benchmark datasets.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2023)
Article
Computer Science, Information Systems
Mohamed Ragab, Emadeldeen Eldele, Wee Ling Tan, Chuan-Sheng Foo, Zhenghua Chen, Min Wu, Chee-Keong Kwoh, Xiaoli Li
Summary: Unsupervised domain adaptation methods aim to generalize well on unlabeled test data with shifted distribution. Existing works on time series domain adaptation suffer from inconsistencies in evaluation schemes, datasets, and neural network architectures. To address these issues, a benchmarking evaluation suite (AdaTime) is developed to evaluate different domain adaptation methods on time series data. Extensive experiments have been conducted to evaluate 11 methods on five datasets, revealing practical insights and building a foundation for future works.
ACM TRANSACTIONS ON KNOWLEDGE DISCOVERY FROM DATA
(2023)
Article
Automation & Control Systems
Xinyu Liu, Xiren Miao, Hao Jiang, Jing Chen, Min Wu, Zhenghua Chen
Summary: This study presents a graph-based relation guided network for power line component detection, which exploits correlations of regions, images, and categories. Experimental results demonstrate that the proposed method can achieve more accurate and reasonable component detection compared to previous methods, which verifies the effectiveness of the proposed model incorporated with relation knowledge.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Xiaoqi Han, Ru Li, Xiaoli Li, Jeff Z. Pan
Summary: This paper proposes a novel framework to address the challenge of correcting errors in language models through divide-and-conquer edits with parallel Editors. Research findings reveal that existing methods often ignore conflicts in multi-edits, whereas our approach can learn diverse editing strategies, resulting in better adaptation to multiple edits.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Computer Science, Information Systems
Zi-Chao Zhang, Xingzhong Zhao, Guiying Dong, Xing-Ming Zhao
Summary: By employing a deep learning model called 3-dimensional multi-task multi-layer perceptron mixer, we have successfully predicted the standardized uptake value ratios (SUVRs) for FDG-PET and AV45-PET from structural magnetic resonance imaging data, which can be used for Alzheimer's disease diagnosis. The proposed method outperforms other methods in terms of accuracy and generalizability, achieving high sensitivity and distinct longitudinal patterns for different disease status.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Automation & Control Systems
Anushiya Arunan, Yan Qin, Xiaoli Li, Chau Yuen
Summary: Data-driven industrial health prognostics benefit from the decentralized and privacy-preserving learning technique of federated learning (FL), which allows utilization of edge industrial data while complying with data privacy laws. However, developing accurate federated models for FL-based health prognostics tasks is challenging due to data heterogeneity among edge devices. We propose a pioneering FL-based health prognostic model with a feature similarity-matched parameter aggregation algorithm to effectively learn from heterogeneous edge data.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2023)
Article
Computer Science, Hardware & Architecture
Devki Nandan Jha, Zhenghua Chen, Shudong Liu, Min Wu, Jiahan Zhang, Graham Morgan, Rajiv Ranjan, Xiaoli Li
Summary: Personalised health and fitness provide users with information regarding their wellbeing and an opportunity to inform healthcare services for better patient outcomes. AE-HAR is a model that balances battery depletion, data accuracy, and timely delivery of results in human activity recognition (HAR). It incorporates a lightweight machine learning component and cloud-based calculations to achieve high accuracy and energy consumption savings.
IEEE TRANSACTIONS ON SUSTAINABLE COMPUTING
(2023)