Article
Computer Science, Artificial Intelligence
Yun Wang, Zhenbo Li, Fei Li, Yang Mi, Jun Yue
Summary: The paper proposes a novel fuzzy discriminative block representation learning algorithm for image feature extraction. By designing effective constraints to enhance the discriminability of the subspace and introducing a transformation matrix for joint optimization, the discriminative ability is further improved.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Computer Science, Artificial Intelligence
Dongmei Mo, Zhihui Lai, Jie Zhou, Hu Qinghua
Summary: In this paper, a new method called Jointly Sparse Orthogonal Linear Discriminant Analysis (JSOLDA) is proposed to improve the performance of linear discriminant analysis in the field of computer vision and pattern recognition. The proposed method obtains the sparse orthogonal projections for feature extraction through constrained scatter matrix decomposition. Experimental results demonstrate that JSOLDA outperforms several well-known methods based on linear discriminant analysis and L2,1-norm.
PATTERN RECOGNITION
(2023)
Article
Engineering, Electrical & Electronic
Zhihui Lai, Zhuozhen Yu, Heng Kong, Linlin Shen
Summary: The paper introduces an improved method for image-based feature extraction called Two-dimensional jointly sparse RDR (2DJSRDR), which utilizes the two-dimensional image matrix directly as the computational unit to avoid drawbacks and improve model performance in previous methods.
SIGNAL PROCESSING-IMAGE COMMUNICATION
(2021)
Article
Computer Science, Information Systems
Hui Huang, Hai-Jun Rong, Zhao-Xu Yang, Chi-Man Vong
Summary: In this paper, a new method VSRP-AnYa-EFS is proposed to address the drawbacks of existing EFSs by combining VSRP and local learning, which results in a compact structure and fast learning speed. Numerical examples demonstrate that this method significantly reduces training time and improves accuracy.
INFORMATION SCIENCES
(2021)
Article
Computer Science, Artificial Intelligence
Yiling Lin, Zhihui Lai, Jie Zhou, Jiajun Wen, Heng Kong
Summary: Researchers propose a generalized robust multiview discriminant analysis method that achieves strong robustness and joint sparsity by reconstructing scatter terms and using L-2,L- 1 regularization. Experimental results demonstrate its significant performance in multiview tasks.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Hardware & Architecture
Xi Chen, Zhihui Lai
Summary: JS-LSSVM algorithm introduces L-2, L-1 norm into LS-SVM, achieving feature selection and dimensionality reduction through projection matrix for better classification performance.
COMPUTERS & ELECTRICAL ENGINEERING
(2022)
Article
Computer Science, Artificial Intelligence
Fei Wang, Quan Wang, Feiping Nie, Zhongheng Li, Weizhong Yu, Rong Wang
Summary: In this paper, Un-LDA is proposed as an extension of LDA for unsupervised subspace learning and clustering. By optimizing the clusters using K-means and the subspace using supervised LDA methods alternately, Un-LDA overcomes the difficulty in solving non-convex objective optimization. Experiments show that Un-LDA algorithms are comparable or superior to existing methods.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2021)
Article
Computer Science, Artificial Intelligence
Xiaowei Zhao, Feiping Nie, Rong Wang, Xuelong Li
Summary: This paper proposes a novel fuzzy K-Means clustering model for conducting clustering tasks on a flexible manifold. The model performs fuzzy clustering based on shrunk patterns with desired manifold structure, and integrates the learning of shrunk patterns and the learning of membership degree into a unified framework. Experimental results demonstrate the feasibility and effectiveness of the proposed clustering algorithms.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Mathematics
Xueyu Chen, Minghua Wan, Hao Zheng, Chao Xu, Chengli Sun, Zizhu Fan
Summary: This paper proposes a new hashing algorithm, BNDDH, which extracts image features by constructing neighborhood graphs and using bilinear projection. Experimental results demonstrate that the proposed algorithm outperforms traditional algorithms in terms of different feature types.
Article
Automation & Control Systems
Zhaoyin Shi, Long Chen, Guang-Yong Chen, Kai Zhao, C. L. Philip Chen
Summary: This study proposes a soft partition clustering method based on the construction of subspaces, which achieves promising performance in image clustering. By using bilinear orthogonal subspaces and calculating reconstruction errors in these subspaces, the clustering of samples is achieved. In addition, graph regularization is applied to preserve the local relational or manifold information of the image data. The proposed method is a one-stage clustering model that reduces the computational burden and maintains the structural relationship between pixels in the image.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Saptarshi Chakraborty, Swagatam Das
Summary: In this paper, a simple and efficient sparse clustering algorithm called LW-k-means is proposed for high-dimensional data. The algorithm incorporates feature weighting to enable feature selection and has a time complexity similar to traditional algorithms. The strong consistency of the LW-k-means procedure is also established. Experimental results on synthetic and real-life datasets demonstrate that LW-k-means performs competitively in terms of clustering accuracy and computational time compared to existing methods for center-based high-dimensional clustering.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Wei Zhang, Zhaohong Deng, Te Zhang, Kup-Sze Choi, Shitong Wang
Summary: This paper proposes a novel method called OMFC-CS for multiview fuzzy clustering, which addresses two challenges by collaborative learning between the common and specific space information. For the first challenge, a mechanism based on matrix factorization is proposed to extract the common and specific information simultaneously. For the second challenge, a one-step learning framework is designed to integrate the learning of common and specific spaces and the learning of fuzzy partitions.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Automation & Control Systems
Yun Wang, Zhenbo Li, Fei Li, Pu Yang, Jun Yue
Summary: In this work, we propose a novel fuzzy discriminative projection and representation learning (FDPR) method for image classification. The method enhances the robustness of the algorithm by designing a fuzzy weight matrix, introducing sparse constraint, and low-rank and l(2,1) norm constraints. Experimental results show that the proposed method performs well on multiple datasets.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2022)
Article
Automation & Control Systems
Runmei Li, Zherui Zhong, Jin Chai, Jian Wang
Summary: This paper proposes a CC-LSTM vehicle trajectory prediction model that combines clustering analysis and feature fusion, which can meet the real-time and accuracy requirements of autonomous vehicles in complex driving environments.
INTERNATIONAL JOURNAL OF FUZZY SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Xin Guan, Yoshikazu Terada
Summary: In this paper, a novel sparse kernel k-means clustering method is proposed to address the issue of clustering high-dimensional data. By optimizing the feature indicators, the clustering performance is improved.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Artificial Intelligence
Ziming Luo, Can Gao, Jie Zhou
Summary: This study proposes a rough sets-based tri-trade model for partially labeled data. A new discernibility matrix is first proposed to consider both labeled and unlabeled data, and a beam search-based algorithm is provided to generate multiple semi-supervised reducts. Then, a tri-trade model is developed using three diverse semi-supervised reducts, with a data editing technique embedded to generate reliable pseudo-labels for unlabeled data. Theoretical analysis and comparative experiments on UCI datasets demonstrate that the proposed model effectively utilize unlabeled data to improve generalization performance and outperform other representative methods.
APPLIED INTELLIGENCE
(2023)
Article
Automation & Control Systems
Jianglin Lu, Jie Zhou, Yudong Chen, Witold Pedrycz, Kwok-Wai Hung
Summary: This article introduces a generalized image transfer retrieval (GITR) problem and proposes an asymmetric transfer hashing (ATH) framework to address it. The ATH framework optimizes asymmetric hash functions and a bipartite graph to achieve knowledge transfer and feature alignment. Experimental results show the superiority of our ATH method in comparison with state-of-the-art hashing methods.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Automation & Control Systems
Zhihui Lai, Xi Chen, Junhong Zhang, Heng Kong, Jiajun Wen
Summary: This article presents a novel framework that integrates discriminative feature extraction and sparse feature selection into the support vector machine, aiming to find the optimal/maximal classification margin. The proposed algorithm, called the maximal margin SVM (MSVM), utilizes an alternatively iterative learning strategy to learn the optimal discriminative sparse subspace and the corresponding support vectors. Experimental results demonstrate the superiority of MSVM over classical discriminant analysis methods and SVM-related methods.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Computer Science, Artificial Intelligence
Jintao Luo, Can Gao, Da Wan, Linlin Shen
Summary: This method proposes a novel unsupervised anomaly detection method, which extracts discriminative features from images using a deep pre-trained network and aggregates shallow and deep features into texture and semantic modules. Then, a feature fusion module is developed to interactively enable feature information in two different modules. Finally, an anomaly segmentation module is designed to generate anomaly detection results by integrating the results of the texture and semantic modules by setting a threshold.
IET COMPUTER VISION
(2023)
Article
Computer Science, Artificial Intelligence
Xiaofeng Tan, Can Gao, Jie Zhou, Jiajun Wen
Summary: In this study, a three-way decision-based co-detection model for unsupervised outlier detection is proposed, which improves the accuracy of the measure of local reachability density by introducing Gaussian kernel function to the local outlier factor (LOF) method and reduces the negative effect of irrelevant and redundant attributes on the measure of sample similarity by using fuzzy rough sets for attribute reduction. A co-detection model trained on both the original view and the transformed view generated by principal component analysis is developed, which collaboratively detects outliers using the strategy of the three-way decision. Comparative experiments on selected UCI datasets show that the proposed model outperforms state-of-the-art methods in terms of AUC-ROC index.
INTERNATIONAL JOURNAL OF APPROXIMATE REASONING
(2023)
Article
Automation & Control Systems
Jun Wan, Hui Xi, Jie Zhou, Zhihui Lai, Witold Pedrycz, Xu Wang, Hang Sun
Summary: This article proposes a self-calibrated pose attention network (SCPAN) framework for more robust and precise facial landmark detection in challenging scenarios. By integrating BALI fields and SCPA model, more facial prior knowledge can be learned and detection accuracy and robustness are improved.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Computer Science, Information Systems
Zhihui Lai, Yiling Lin, Jiacan Zheng, Jie Zhou, Heng Kong
Summary: In this paper, a novel multiview learning framework called generalized multiview regression (GMR) is proposed to address the limitations of CCA-based methods and improve the performance of multiview subspace learning. GMR aims to find a common subspace that preserves the complementary information of each view and maintains consistency among all the views. It incorporates data reconstruction, local geometric structures, and an orthogonal dictionary to capture discriminative consistency. The use of L2,1 as the basic norm facilitates robustness and sparsity for feature extraction and selection.
INFORMATION SCIENCES
(2023)
Article
Mathematics
Liuxin Wang, Can Gao, Jie Zhou, Jiajun Wen
Summary: In this study, a semi-supervised co-training model based on the three-way decision and pseudo labels is proposed for classifying partially labeled data. A simple yet effective method is used to assign pseudo labels to unlabeled data, and a heuristic attribute reduction algorithm is developed. The three-way decision is combined with entropy to form co-decision rules for classifying unlabeled data, and useful samples are iteratively selected to improve the co-decision model's performance. Experimental results on UCI datasets demonstrate the proposed model's potential for partially labeled data.
Article
Computer Science, Artificial Intelligence
Weilin Huang, Zhihui Lai, Heng Kong, Junhong Zhang
Summary: This article proposes a novel method called Joint Sparse Locality Preserving Regression (JSLPR) for discriminative learning. JSLPR considers the local geometric structure of the data and uses L-2, L-1 norms for robustness and feature selection. Experimental results demonstrate the superiority of JSLPR in feature extraction and selection.
IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE
(2023)
Article
Computer Science, Information Systems
Jianjun Qian, Shumin Zhu, Chaoyu Zhao, Jian Yang, Wai Keung Wong
Summary: This paper proposes a hard samples guided optimal transport (OT) loss, OTFace, to improve face representation in the wild. It enhances the performance of hard samples by introducing feature distribution discrepancy while maintaining the performance on easy samples.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Computer Science, Artificial Intelligence
Jun Wan, Jun Liu, Jie Zhou, Zhihui Lai, Linlin Shen, Hang Sun, Ping Xiong, Wenwen Min
Summary: Facial landmark detection methods often predict landmarks by mapping input facial appearance features to landmark heatmaps, but they struggle with large poses, occlusions, and varied illuminations. To address this, we propose a novel Reference Heatmap Transformer (RHT) that utilizes reference heatmap information for more accurate detection of facial landmarks. Experimental results on benchmark datasets show that our method outperforms existing state-of-the-art techniques.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Yuwu Lu, Wai Keung Wong, Biqing Zeng, Zhihui Lai, Xuelong Li
Summary: This paper proposes a guided discrimination and correlation subspace learning (GDCSL) method for cross-domain image classification. GDCSL considers the domain-invariant, category-discriminative, and correlation learning of data. It introduces the discriminative information associated with the source and target data and extracts the most correlated features for image classification. Experimental results show the effectiveness of the proposed methods over state-of-the-art domain adaptation methods.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Jie Zhou, Can Gao, Xizhao Wang, Zhihui Lai, Jun Wan, Xiaodong Yue
Summary: In this study, the notion of typicality-aware adaptive similarity matrix learning is presented to address the issues of unreasonable similarity matrix construction and sensitivity to noisy scenarios. By measuring and adaptively learning the typicality (possibility) of each sample being a neighbor of other samples, the impact of noisy data or outliers can be alleviated and the neighborhood structures can be well captured. Moreover, the generated similarity matrix has beneficial block diagonal properties for correct clustering.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Xu Wu, Zhihui Lai, Shiqi Yu, Jie Zhou, Zhuoqian Liang, Linlin Shen
Summary: Low-light image enhancement aims to improve illumination intensity and restore color information. To overcome the limitations of existing methods, researchers propose a novel pipeline, called LRCR-Net, which performs light restoration and color refinement in a coarse-to-fine manner. The proposed pipeline incorporates region-calibrated residual block and learnable image processing operators to achieve better performance in complex lighting scenes and dark regions.
IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE
(2023)