Article
Computer Science, Information Systems
Bo Liu, Xiaodong Chen, Yanshan Xiao, Weibin Li, Laiwang Liu, Changdong Liu
Summary: Multi-view learning explores information from different perspectives, with dictionary learning being advantageous for classification, but there is limited research on combining the two. The proposed MVDL-CV method enhances multi-view classification by learning specific dictionaries and utilizing regularization between them for improved discriminative representation.
INFORMATION SCIENCES
(2021)
Article
Biochemical Research Methods
Hoon Seo, Lodewijk Brand, Lucia Saldana Barco, Hua Wang
Summary: In this study, a novel Primal-Dual Multi-Instance Support Vector Machine method is proposed for automating the histopathological classification of breast cancer. By bypassing common optimization methods, the proposed method is computationally efficient and achieves promising results in experiments.
Article
Computer Science, Artificial Intelligence
Shiping Wang, Zhewen Wang, Kart-Leong Lim, Guobao Xiao, Wenzhong Guo
Summary: This paper proposes a simple and efficient supervised random walk scheme for addressing the multi-view semi-supervised classification problem, which demonstrates superiority compared to other methods. Additionally, the proposed method shows positive robustness and promising generalization capability in terms of data classification.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Wanrong Yu, Xiao-Jun Wu, Tianyang Xu, Ziheng Chen, Josef Kittler
Summary: SC exploits the potential capacity of self-expressive modeling and constructs an affinity matrix. However, a strict affine constraint is not flexible enough to handle real-world cases. Therefore, a scalable affine constraint is introduced to enhance clustering performance.
NEURAL PROCESSING LETTERS
(2023)
Article
Computer Science, Artificial Intelligence
Sheng Wu, Ancong Wu, Wei-Shi Zheng
Summary: In this work, we propose an online multi-view learning approach that utilizes the principles of multi-view complementarity and consistency to effectively process online multi-view data. Diverse features extracted from different deep feature extractors under different views are used as input to an online learning method for the discovery and memorization of view-specific information. The proposed approach includes strategies such as a softmax-weighted reducible (SWR) loss for selective view fusion, cross-view embedding consistency (CVEC) loss and cross-view Kullback-Leibler (CVKL) divergence loss for maintaining cross-view consistency, and a knowledge registration unit (KRU) based on dictionary learning to handle knowledge forgetting.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Arthur Hoarau, Arnaud Martin, Jean-Christophe Dubois, Yolande Le Gall
Summary: This paper proposes an Evidential Decision Tree and an Evidential Random Forest, which can handle uncertain and imprecise predictions and can predict rich labels. Experimental results showed better performance for the presented methods compared to other evidential models and recent Cautious Random Forests in handling noisy data and effectively uncertainly and imprecisely labeled datasets. The proposed models also offer better robustness and the ability to predict rich labels, which can be used in other approaches such as active learning.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Xin Zheng, Shouzhi Liang, Bo Liu, Xiaoming Xiong, Xianghong Hu, Yuan Liu
Summary: The paper proposes a new architecture named GMADL for subgraph feature extraction, utilizing dictionary learning approaches to enhance discrimination of model features in graph data. By designing an analysis dictionary and constructing multi-view support vector machine classifiers, the efficiency of feature extraction is improved and the classification model prediction accuracy is enhanced by utilizing information from multiple views. Comparisons with state-of-the-art approaches demonstrate the feasibility and competitiveness of the proposed architecture in graph classification.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Engineering, Mechanical
Shenqing Xiao, Mingliang Li, Bo Chen, Xingye Zhou, Chenchen Xi, Yiqiu Tan
Summary: To understand the evolution of pavement texture, the self-affine characteristics and spatiotemporal variation of pavement texture were analyzed using long-term field observation data. The texture roughness of the same pavement shows similarity at a small scale but variability at a large scale. This variability is also reflected in the spatial variation of on-site texture depth. Additionally, the pavement texture depth increases over time, especially with traffic polishing, which leads to enlarged particle gaps and smoother particle surfaces.
TRIBOLOGY INTERNATIONAL
(2023)
Article
Computer Science, Artificial Intelligence
Saroj S. Shivagunde, Ashwani Nadapana, V. Vijaya Saradhi
Summary: The paper proposes Multi-view Incremental Discriminant Analysis (MvIDA) algorithm, which can update the trained model in incremental data. Experimental results show that MvIDA outperforms batch multi-view algorithms in terms of order independence and faster construction of optimal discriminant subspace.
INFORMATION FUSION
(2021)
Article
Computer Science, Information Systems
Saroj S. Shivagunde, V. Vijaya Saradhi
Summary: With the advancement of technology, the size of high-definition images has grown exponentially, making traditional 1D methods less suitable for handling such data. To address this issue, we propose a generalized method called 2D Multi-view Discriminant Analysis (2DMvDA), which directly uses 2D image matrices for classification, resulting in a significant reduction in data size.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Information Systems
E. Rachdi, I. El Khadiri, Y. El merabet, Y. Rhazi, C. Meurie
Summary: This paper introduces a novel local feature extraction operator called MTSP, which is composed of two single-scale encoders, STP and SSP, designed based on a novel set theory pattern encoding scheme. Unlike other parametric texture operators, MTSP incorporates dynamic thresholds and can capture more detailed image information through the fusion of STP and SSP encoders. Experimental results demonstrate that MTSP achieves reliable performance stability on ten texture datasets and outperforms several representative methods in texture modeling, as verified by statistical tests.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2023)
Article
Biochemistry & Molecular Biology
Gan Wang, Xudong Zhang, Zheng Pan, Alfonso Rodriguez Paton, Shuang Wang, Tao Song, Yuanqiang Gu
Summary: A novel end-to-end learning model was proposed to encode drugs and proteins and was evaluated on the BindingDB dataset, outperforming other models.
Article
Computer Science, Artificial Intelligence
Wei Guo, Zhe Wang, Ziqiu Chi, Xinlei Xu, Dongdong Li, Songyang Wu
Summary: In this paper, we propose a one-stage multi-view subspace clustering with dictionary learning (OSMvSC) method to address the problem of clustering heterogeneous features in multi-view communities. The proposed method integrates dictionary learning, representation coefficient matrix learning, and matrix factorization to learn the original multi-view data and achieve clustering results with linear time complexity. Experimental results on various real-world multi-view datasets demonstrate the effectiveness and superiority of the proposed method.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Steve Tsham Mpinda Ataky, Alessandro Lameiras Koerich
Summary: Texture descriptors are widely used in medical image analysis, especially in histopathologic images (HIs), due to the variability of texture and tissue appearance caused by staining irregularities. However, extracting texture features in a discriminant way is challenging because of the non-deterministic complex system formed by the intrinsic properties of such images. This paper proposes a novel method that quantifies these properties using ecological diversity measures and discrete wavelet transform, which outperforms state-of-the-art shallow and deep methods according to experimental results on two HI datasets.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Mashaan Alshammari, John Stavrakakis, Adel F. Ahmed, Masahiro Takatsuka
Summary: Partitioning trees are efficient for k-nearest neighbor search, but kd-trees can be ineffective in high dimensions. Random projection trees (rpTrees) solve this problem and are influenced by point dispersion and the number of rpTrees in an rpForest.
Article
Radiology, Nuclear Medicine & Medical Imaging
Yunpeng Dai, Duofang Chen, Guodong Wang, Jipeng Yin, Yonghua Zhan, Kaichun Wu, Jimin Liang, Xueli Chen
QUANTITATIVE IMAGING IN MEDICINE AND SURGERY
(2020)
Article
Oncology
Ke Li, Yun Zhou, Yulong Chen, Liansuo Zhou, Jimin Liang
CANCER CHEMOTHERAPY AND PHARMACOLOGY
(2020)
Article
Neurosciences
Yue Wang, Jianpu Yan, Zhongliang Yin, Shenghan Ren, Minghao Dong, Changli Zheng, Wei Zhang, Jimin Liang
Summary: Visual processing involves perceiving, analyzing, and synthesizing visual objects, with neural activities extending from visual cortex to higher cognitive areas. Studies have shown that complex backgrounds and divided attention can decrease object detection accuracy and prolong decision time. Research on scene complexity and attentional state can help understand brain responses in different experimental conditions.
FRONTIERS IN NEUROSCIENCE
(2021)
Article
Neurosciences
Yue Wang, Chenwang Jin, Zhongliang Yin, Hongmei Wang, Ming Ji, Minghao Dong, Jimin Liang
Summary: Visual expertise is attributable to accumulated experience in a specific domain, leading to widespread neural activities beyond the visual cortex. Studying a group of radiological interns, researchers found that visual experience can modulate brain connectivity, enhancing integration within visual processing circuits and between high-order brain circuits. This controlled and interactive process is influenced by multiple top-down factors and can enhance participants' acquisition of specific visual information.
HUMAN BRAIN MAPPING
(2021)
Article
Computer Science, Artificial Intelligence
Chen Wei, Yiping Tang, Chuang Niu Chuang Niu, Haihong Hu, Yue Wang, Jimin Liang
Summary: This study introduces a new neural predictor architecture encoding scheme and two self-supervised learning methods to improve the predictive performance of neural predictors. Experimental results demonstrate that these methods can achieve comparable or superior performance with only half of the training samples compared to supervised counterparts. Applying these methods to the NPENAS algorithm results in state-of-the-art performance on NASBench-101, NASBench-201, and DARTS benchmarks.
IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE
(2021)
Article
Computer Science, Interdisciplinary Applications
Qiuju Yang, Chang Liu, Jimin Liang
Summary: The study introduces an auroral image clustering network (AICNet), which can automatically classify all-sky images and discover the internal structures of auroras, significantly improving the efficiency of auroral morphology classification.
EARTH SCIENCE INFORMATICS
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Zhenzhen Xu, Bo Tao, Chuanbin Liu, Dong Han, Jibin Zhang, Junsong Liu, Sulei Li, Weijie Li, Jing Wang, Jimin Liang, Feng Cao
Summary: The novel 3D fusion quantitative assessment method utilizing CTA and SPECT images provides reliable and intuitive evaluations of left ventricular infarction. The method was experimentally validated and showed correlation with histological analysis and clinical perfusion SPECT results, demonstrating its potential for pre-operation evaluation and post-diagnosis management of myocardial infarction patients.
QUANTITATIVE IMAGING IN MEDICINE AND SURGERY
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Duofang Chen, Zhaohui Wang, Kai Chen, Qi Zeng, Lin Wang, Xinyi Xu, Jimin Liang, Xueli Chen
Summary: The combination of lensless digital holography and machine learning, specifically convolutional neural networks, shows promising results in effectively classifying tumor cells, which is crucial for cancer diagnosis and research.
QUANTITATIVE IMAGING IN MEDICINE AND SURGERY
(2021)
Article
Geochemistry & Geophysics
Yiping Tang, Kaitai Guo, Chen Wei, Yang Zheng, Shenghan Ren, Jimin Liang
Summary: PMAFs, common dayside auroral forms, can be recognized and tracked using a poleward motion aware network (PA-Net). PA-Net avoids complicated optical flow estimation, enabling application on large-scale auroral datasets.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Engineering, Electrical & Electronic
Lei Hu, Chuang Niu, Shenghan Ren, Minghao Dong, Changli Zheng, Wei Zhang, Jimin Liang
Summary: This article introduces a target extraction neural network named "discriminative context-aware network," focusing on capturing rich context information and preserving spatial information, achieving remarkable results.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2022)
Article
Computer Science, Artificial Intelligence
Zhenzhen You, Ming Jiang, Zhenghao Shi, Cheng Shi, Shuangli Du, Jimin Liang, Anne-Sophie Herard, Caroline Jan, Nicolas Souedet, Thierry Delzescaux
Summary: In this study, a method combining multiscale fully convolutional regression neural network and competitive region growing technique successfully achieved individualization of size-varying and touching neurons in major anatomical regions of the macaque brain.
NEURAL COMPUTING & APPLICATIONS
(2022)
Article
Engineering, Biomedical
Getao Du, Yonghua Zhan, Yue Zhang, Jianzhong Guo, Xueli Chen, Jimin Liang, Heng Zhao
Summary: This study proposes an improved automatic segmentation method based on a U-net network to segment the gastrocnemius and soleus muscles in ultrasound images of the shank. Experimental results demonstrate that the method achieves superior segmentation capability and accuracy.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Huiyuan Wang, Nan Wang, Hui Xie, Lin Wang, Wangting Zhou, Defu Yang, Xu Cao, Shouping Zhu, Jimin Liang, Xueli Chen
Summary: A two-stage deep learning network-based framework was proposed for reconstruction using few-view projections, exhibiting strong abilities in image reconstruction and showing great potential in in vivo PT applications.
QUANTITATIVE IMAGING IN MEDICINE AND SURGERY
(2022)
Article
Chemistry, Analytical
Chen Wei, Shenghan Ren, Kaitai Guo, Haihong Hu, Jimin Liang
Summary: In this study, a high-resolution Swin Transformer network (HRSTNet) is proposed for medical image segmentation. The HRSTNet replaces convolutional layers with Transformer blocks and continuously exchanges feature map information with different resolutions. Experimental results demonstrate that the HRSTNet achieves comparable performance with the state-of-the-art Transformer-based U-Net-like architecture on multiple medical image datasets.
Article
Computer Science, Artificial Intelligence
Chen Wei, Chuang Niu, Yiping Tang, Yue Wang, Haihong Hu, Jimin Liang
Summary: This article introduces a neural predictor guided evolutionary algorithm to enhance the exploration ability of neural architecture search (NAS) and designs two types of neural predictors. Experimental results indicate that this method achieves good performance in NAS tasks.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)