Article
Computer Science, Artificial Intelligence
Rui Wang, Xiao-Jun Wu, Tianyang Xu, Cong Hu, Josef Kittler
Summary: This paper proposes a U-shaped neural network (U-SPDNet) based on SPD manifolds for visual classification. The U-SPDNet consists of an encoder and a decoder to extract and reconstruct image features, respectively, and addresses the degradation of structural information. Additionally, skip connections and geometric operations are employed to enhance the representational capacity of U-SPDNet, resulting in improved accuracy on multiple datasets.
Article
Computer Science, Artificial Intelligence
Byung Hyung Kim, Jin Woo Choi, Honggu Lee, Sungho Jo
Summary: This paper proposes a new Riemannian-based deep learning network for generating more discriminative features in EEG classification. The model learns the Riemannian barycenter for each class in a Riemannian geometric space and normalizes the distribution of SPD matrices. Experimental results demonstrate the superiority of the proposed framework in learning the non-stationary nature of EEG signals.
PATTERN RECOGNITION
(2023)
Article
Automation & Control Systems
Xiaofeng Liao, You Zhao, Xian Zhou
Summary: In recent years, Riemannian geometry has been widely applied to solve nonlinear programming problems, and a class of neurodynamic flow approaches has been proposed, which have global convergence and feasibility in seeking the minimum point of convex and quasi-convex minimum problems. Furthermore, the approach can also be adapted to competitive neural networks and image and signal processing in compressive sensing.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2023)
Article
Computer Science, Information Systems
Ziheng Chen, Tianyang Xu, Xiao-Jun Wu, Rui Wang, Josef Kittler
Summary: With the increasing amount of video data, image set classification has become a popular topic in the field of computer vision and pattern recognition. However, the diversity within classes and ambiguity between classes pose a challenge. To address this, multiple geometry-aware image set modelling and learning methods have been proposed. In this paper, we propose a hybrid Riemannian metric learning framework that effectively fuses complementary kernel features obtained from different manifolds into a unified subspace for classification. Our approach achieves improved efficiency and outperforms state-of-the-art methods according to experimental results.
IEEE TRANSACTIONS ON BIG DATA
(2023)
Article
Computer Science, Information Systems
Huiyuan Deng, Xiangzhu Meng, Huibing Wang, Lin Feng
Summary: In this paper, we propose an Optimal instance Partition-based Multi-Metric Learning (OPM2L) method for heterogeneous dataset classification. The method unifies instance partition and multiple local metrics learning into a single objective, improving the learning process and showing better performance through experimental results.
INFORMATION SCIENCES
(2023)
Article
Computer Science, Information Systems
Rui Wang, Xiao-Jun Wu, Kai-Xuan Chen, Josef Kittler
Summary: This study discusses the importance of recognizing wild video based image sets and proposes a novel algorithm to model image sets from a multi-geometric perspective for improved classification performance.
IEEE TRANSACTIONS ON BIG DATA
(2022)
Article
Biology
Yunyuan Gao, Xinyu Sun, Ming Meng, Yingchun Zhang
Summary: This research utilizes Riemannian geometry for EEG-based emotion recognition, extracting time-frequency features to construct SPD matrices, designing a dimensionality reduction algorithm, and employing multiple classification methods to achieve high recognition accuracy.
COMPUTERS IN BIOLOGY AND MEDICINE
(2022)
Article
Computer Science, Artificial Intelligence
Wanguang Yin, Zhengming Ma, Quanying Liu
Summary: Discriminative subspace learning is an important problem in machine learning. We propose MODA, a manifold optimization-based discriminant analysis method, which achieves the best separability and is significantly superior to competing algorithms. Experimental results on various datasets show that MODA has a higher accuracy, especially for time series of EEG signals. The code for MODA is available at https://github.com/ncclabsustech/MODA-algorithm.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Artificial Intelligence
Dong Wei, Xiaobo Shen, Quansen Sun, Xizhan Gao
Summary: In this paper, a novel Discrete Metric Learning (DML) approach based on the Riemannian manifold is proposed for fast image set classification. It achieves competitive performance and efficiency compared to existing methods.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Computer Science, Artificial Intelligence
Zirui Zhang, Yinan Guo, Fengzhen Tang
Summary: This paper proposes a novel dimension selection method called DSSR, which improves the performance of EEG classification by eliminating redundant dimensions for SPD matrices in the Riemannian space.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Feihu Huang, Shangqian Gao
Summary: The paper investigates novel stochastic composition optimization problems over Riemannian manifolds, proposing RCG and M-RCG algorithms with different sample complexities for solving these problems. Extensive numerical experiments demonstrate the effectiveness of these algorithms in training DNNs and learning PCA. This study represents the first exploration of composition optimization problems over Riemannian manifolds.
Article
Computer Science, Information Systems
Yangyang Li, Ruqian Lu
Summary: Researchers propose a manifold learning assumption called manifold learning (MAL) to describe the nonlinear distribution of an image dataset. By learning the real Riemannian metric, they aim to uncover the intrinsic geometric structure of the embedded manifold. They propose a novel algorithm, curvature flow, which incorporates curvature information into metric learning. Through optimizing the objective function, they obtain a set of iterative equations and a Mahalanobis metric that effectively uncovers the intrinsic structure of the manifold. Experimental results show that their proposed method outperforms other algorithms.
SCIENCE CHINA-INFORMATION SCIENCES
(2022)
Article
Mathematics, Applied
Milan Lj. Zlatanovic, Miroslav D. Maksimovic
Summary: We define and study the quarter-symmetric connection that preserves the generalized metric G in the generalized Riemannian manifold. It is proven that the skew-symmetric part F of the generalized metric G in the manifold with the quarter-symmetric generalized metric connection is closed, implying that an even-dimensional manifold is a symplectic manifold. We also explore the properties of curvature tensors and connection transformations, where the Riemannian tensor of the Levi-Civita connection remains invariant. Finally, we observe the quarter-symmetric connection under special conditions.
Article
Computer Science, Software Engineering
Ke Ye, Ken Sze-Wai Wong, Lek-Heng Lim
Summary: The article presents tools for optimizing over a set of flags, which is a smooth manifold known as the flag manifold, including the Grassmannian as a special case. Various differential geometric objects are derived with closed-form analytic expressions for Riemannian optimization algorithms on the flag manifold, introducing systems of extrinsic coordinates to parameterize points, metrics, tangent spaces, geodesics, distances, parallel transports, gradients, Hessians in terms of matrices and matrix operations, allowing for the formulation of steepest descent, conjugate gradient, and Newton algorithms using standard numerical linear algebra.
MATHEMATICAL PROGRAMMING
(2022)
Article
Computer Science, Information Systems
Krishan Sharma, Renu Rameshan
Summary: This paper explores the inherent geometry of video tensors by modeling them as points in product of Riemannian matrix manifolds and proposes positive definite kernels for feature mapping and classification. Experimental results on publicly available datasets show that the proposed methodology outperforms state-of-the-art methods in terms of classification accuracy.
JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION
(2021)
Article
Automation & Control Systems
Donglin Zhang, Xiao-Jun Wu, Tianyang Xu, Josef Kittler
Summary: This paper proposes a novel two-stage supervised discrete hashing (TSDH) method to address the issues in existing cross-media hashing approaches. By generating latent representations and binary codes in a common hash space, and by directly endowing the hash codes with semantic labels and using a discrete hash optimization approach, the discriminative power of learned binary codes can be enhanced.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Zhongwei Shen, Xiao-Jun Wu, Tianyang Xu
Summary: This paper proposes a Foreground EXtraction (FEX) block to disentangle foregrounds from the background for advanced action recognition systems. The FEX block contains a Foreground Enhancement (FE) module and a Scene Segregation (SS) module, which effectively models foreground clues and splits feature maps for action inference.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
Rui Wang, Xiao-Jun Wu, Josef Kittler
Summary: A SymNet network was proposed for image set classification, which utilized SPD matrix mapping layers, rectifying layers, pooling layers, and log-map layer to achieve effective feature learning and data compression. PCA and KDA algorithms were applied for discriminative subspace learning, and extensive experiments validated the feasibility and effectiveness of the proposed SymNet.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Zhenxing Liu, Xiaoning Song, Zhenhua Feng, Tianyang Xu, Xiaojun Wu, Josef Kittler
Summary: This passage discusses the progress and challenges in pedestrian detection research, proposes a method to enhance pedestrian detection by extracting effective features using contextual information, and validates the effectiveness of the proposed method through experimental results on two benchmark datasets.
NEURAL PROCESSING LETTERS
(2023)
Article
Chemistry, Physical
Yafang Hou, Yuqing Wang, Tianyang Xu, Zhi Wang, Weidong Tian, Di Sun, Xinyue Yu, Pengyao Xing, Jinglin Shen, Xia Xin, Jingcheng Hao
Summary: This article demonstrates a multi-bond-induced hierarchical self-assembly method, which utilizes atomically precise silver nanoclusters to achieve ordered layer-by-layer construction of metal-organic frameworks. The luminescence properties can be reversibly switched by tuning the pH values. This method has potential applications in the fields of luminescent devices and sensors.
CHEMISTRY OF MATERIALS
(2022)
Article
Computer Science, Artificial Intelligence
Yingjie Jiang, Xiaoning Song, Tianyang Xu, Zhenhua Feng, Xiaojun Wu, Josef Kittler
Summary: Siamese trackers have become the mainstream framework for visual object tracking in recent years. This paper proposes a target-cognisant Siamese network that enhances the interaction between the classification and regression branches, and introduces attention mechanisms and filtering modules to improve the tracking performance. Experimental results demonstrate the competitiveness of the proposed method.
PATTERN RECOGNITION LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Rui Wang, Xiao-Jun Wu, Zhen Liu, Josef Kittler
Summary: This paper proposes a geometry-aware graph embedding projection metric learning algorithm to address the challenges of intraclass diversity and interclass similarity in image set classification. The algorithm constructs similarity graphs and utilizes local structural information on the Grassmann manifold for graph learning, and formulates the dimensionality reduction problem into a metric learning regularization term.
IEEE TRANSACTIONS ON COGNITIVE AND DEVELOPMENTAL SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Rui Wang, Xiao-Jun Wu, Tianyang Xu, Cong Hu, Josef Kittler
Summary: This paper proposes a U-shaped neural network (U-SPDNet) based on SPD manifolds for visual classification. The U-SPDNet consists of an encoder and a decoder to extract and reconstruct image features, respectively, and addresses the degradation of structural information. Additionally, skip connections and geometric operations are employed to enhance the representational capacity of U-SPDNet, resulting in improved accuracy on multiple datasets.
Article
Computer Science, Information Systems
Ziheng Chen, Tianyang Xu, Xiao-Jun Wu, Rui Wang, Josef Kittler
Summary: With the increasing amount of video data, image set classification has become a popular topic in the field of computer vision and pattern recognition. However, the diversity within classes and ambiguity between classes pose a challenge. To address this, multiple geometry-aware image set modelling and learning methods have been proposed. In this paper, we propose a hybrid Riemannian metric learning framework that effectively fuses complementary kernel features obtained from different manifolds into a unified subspace for classification. Our approach achieves improved efficiency and outperforms state-of-the-art methods according to experimental results.
IEEE TRANSACTIONS ON BIG DATA
(2023)
Article
Computer Science, Artificial Intelligence
Hui Li, Tianyang Xu, Xiao-Jun Wu, Jiwen Lu, Josef Kittler
Summary: Deep learning based fusion methods have achieved promising performance in image fusion tasks due to the importance of network architecture. However, designing fusion networks is still a challenging task. In this paper, the fusion task is mathematically formulated and a connection between the optimal solution and network architecture is established. This leads to the proposal of a lightweight fusion network based on a learnable representation approach.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Ze Kang, Tianyang Xu, Xue-Feng Zhu, Xiao-Jun Wu
Summary: Traditional Siamese networks for visual tracking rely on offline-trained appearance models for each frame, disregarding temporal variation at the online stage. We propose a novel Motion-Perceive Siamese network (SiamMP) that explicitly predicts motion patterns to enhance appearance-only formulation.
PATTERN RECOGNITION LETTERS
(2023)
Article
Computer Science, Artificial Intelligence
Tianyang Xu, Zhenhua Feng, Xiao-Jun Wu, Josef Kittler
Summary: In this study, a novel network with a target-agnostic object detection module is proposed to complement direct target inference and minimize the misalignment of key cues in potential template-instance matches. A cross-task interaction module is developed to ensure consistent supervision of classification and regression branches, improving their synergy. Adaptive labels are assigned to effectively supervise network training. Experimental results on various benchmarks demonstrate the effectiveness of the advanced target detection module and cross-task interaction, outperforming state-of-the-art tracking methods.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Zhe Chen, Xiao-Jun Wu, Tianyang Xu, Josef Kittler
Summary: The DPL-SCSR algorithm proposed in this article utilizes the label matrix of the dictionary to project the representation and approximate a block-diagonal structure by imposing non-negative constraint and controlling scale. It seamlessly integrates a linear classifier and feature extraction process, reducing training and parameter tuning complexity. Experimental results on image classification datasets show its superiority over state-of-the-art dictionary learning methods.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Hamdan Abdellatef, Lina J. Karam
Summary: This paper proposes performing the learning and inference processes in the compressed domain to reduce computational complexity and improve speed of neural networks. Experimental results show that modified ResNet-50 in the compressed domain is 70% faster than traditional spatial-based ResNet-50 while maintaining similar accuracy. Additionally, a preprocessing step with partial encoding is suggested to improve resilience to distortions caused by low-quality encoded images. Training a network with highly compressed data can achieve good classification accuracy with significantly reduced storage requirements.
Article
Computer Science, Artificial Intelligence
Victor R. Barradas, Yasuharu Koike, Nicolas Schweighofer
Summary: Inverse models are essential for human motor learning as they map desired actions to motor commands. The shape of the error surface and the distribution of targets in a task play a crucial role in determining the speed of learning.
Article
Computer Science, Artificial Intelligence
Ting Zhou, Hanshu Yan, Jingfeng Zhang, Lei Liu, Bo Han
Summary: We propose a defense strategy that reduces the success rate of data poisoning attacks in downstream tasks by pre-training a robust foundation model.
Article
Computer Science, Artificial Intelligence
Hao Sun, Li Shen, Qihuang Zhong, Liang Ding, Shixiang Chen, Jingwei Sun, Jing Li, Guangzhong Sun, Dacheng Tao
Summary: In this paper, the convergence rate of AdaSAM in the stochastic non-convex setting is analyzed. Theoretical proof shows that AdaSAM has a linear speedup property and decouples the stochastic gradient steps with the adaptive learning rate and perturbed gradient. Experimental results demonstrate that AdaSAM outperforms other optimizers in terms of performance.
Article
Computer Science, Artificial Intelligence
Juntong Yun, Du Jiang, Li Huang, Bo Tao, Shangchun Liao, Ying Liu, Xin Liu, Gongfa Li, Disi Chen, Baojia Chen
Summary: In this study, a dual manipulator grasping detection model based on the Markov decision process is proposed. By parameterizing the grasping detection model of dual manipulators using a cross entropy convolutional neural network and a full convolutional neural network, stable grasping of complex multiple objects is achieved. Robot grasping experiments were conducted to verify the feasibility and superiority of this method.
Article
Computer Science, Artificial Intelligence
Miaohui Zhang, Kaifang Li, Jianxin Ma, Xile Wang
Summary: This paper proposes an unsupervised person re-identification (Re-ID) method that uses two asymmetric networks to generate pseudo-labels for each other by clustering and updates and optimizes the pseudo-labels through alternate training. It also designs similarity compensation and similarity suppression based on the camera ID of pedestrian images to optimize the similarity measure. Extensive experiments show that the proposed method achieves superior performance compared to state-of-the-art unsupervised person re-identification methods.
Article
Computer Science, Artificial Intelligence
Florian Bacho, Dominique Chu
Summary: This paper proposes a new approach called the Forward Direct Feedback Alignment algorithm for supervised learning in deep neural networks. By combining activity-perturbed forward gradients, direct feedback alignment, and momentum, this method achieves better performance and convergence speed compared to other local alternatives to backpropagation.
Article
Computer Science, Artificial Intelligence
Xiaojian Ding, Yi Li, Shilin Chen
Summary: This research paper addresses the limitations of recursive feature elimination (RFE) and its variants in high-dimensional feature selection tasks. The proposed algorithms, which introduce a novel feature ranking criterion and an optimal feature subset evaluation algorithm, outperform current state-of-the-art methods.
Article
Computer Science, Artificial Intelligence
Naoko Koide-Majima, Shinji Nishimoto, Kei Majima
Summary: Visual images observed by humans can be reconstructed from brain activity, and the visualization of arbitrary natural images from mental imagery has been achieved through an improved method. This study provides a unique tool for directly investigating the subjective contents of the brain.
Article
Computer Science, Artificial Intelligence
Huanjie Tao, Qianyue Duan
Summary: In this paper, a hierarchical attention network with progressive feature fusion is proposed for facial expression recognition (FER), addressing the challenges posed by pose variation, occlusions, and illumination variation. The model achieves enhanced performance by aggregating diverse features and progressively enhancing discriminative features.
Article
Computer Science, Artificial Intelligence
Zhenyi Wang, Pengfei Yang, Linwei Hu, Bowen Zhang, Chengmin Lin, Wenkai Lv, Quan Wang
Summary: In the face of the complex landscape of deep learning, we propose a novel subgraph-level performance prediction method called SLAPP, which combines graph and operator features through an innovative graph neural network called EAGAT, providing accurate performance predictions. In addition, we introduce a mixed loss design with dynamic weight adjustment to improve predictive accuracy.
Article
Computer Science, Artificial Intelligence
Yiyang Yin, Shuangling Luo, Jun Zhou, Liang Kang, Calvin Yu-Chian Chen
Summary: Medical image segmentation is crucial for modern healthcare systems, especially in reducing surgical risks and planning treatments. Transanal total mesorectal excision (TaTME) has become an important method for treating colon and rectum cancers. Real-time instance segmentation during TaTME surgeries can assist surgeons in minimizing risks. However, the dynamic variations in TaTME images pose challenges for accurate instance segmentation.
Article
Computer Science, Artificial Intelligence
Teng Cheng, Lei Sun, Junning Zhang, Jinling Wang, Zhanyang Wei
Summary: This study proposes a scheme that combines the start-stop point signal features for wideband multi-signal detection, called Fast Spectrum-Size Self-Training network (FSSNet). By utilizing start-stop points to build the signal model, this method successfully solves the difficulty of existing deep learning methods in detecting discontinuous signals and achieves satisfactory detection speed.
Article
Computer Science, Artificial Intelligence
Wenming Wu, Xiaoke Ma, Quan Wang, Maoguo Gong, Quanxue Gao
Summary: The layer-specific modules in multi-layer networks are critical for understanding the structure and function of the system. However, existing methods fail to accurately characterize and balance the connectivity and specificity of these modules. To address this issue, a joint learning graph clustering algorithm (DRDF) is proposed, which learns the deep representation and discriminative features of the multi-layer network, and balances the connectivity and specificity of the layer-specific modules through joint learning.
Article
Computer Science, Artificial Intelligence
Guanghui Yue, Guibin Zhuo, Weiqing Yan, Tianwei Zhou, Chang Tang, Peng Yang, Tianfu Wang
Summary: This paper proposes a novel boundary uncertainty aware network (BUNet) for precise and robust colorectal polyp segmentation. BUNet utilizes a pyramid vision transformer encoder to learn multi-scale features and incorporates a boundary exploration module (BEM) and a boundary uncertainty aware module (BUM) to handle boundary areas. Experimental results demonstrate that BUNet outperforms other methods in terms of performance and generalization ability.