检索结果-内蒙古大学图书馆

Sketching Data Distribution by Rotation

IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING 2024年第11期36卷 7015-7029页

作者： Lei, Runze Wang, Pinghui Li, Rundong Jia, Peng Zhao, Junzhou Guan, Xiaohong Xi An Jiao Tong Univ MOE Key Lab Intelligent Networks & Network Secur Xian 710049 Shaanxi Peoples R China

Kernel density estimation is a useful method for estimating the probability distribution of data. It is a challenge to achieve efficient kernel density estimation, especially for large-scale and high-dimension stream data. We propose rotation kernel, , a novel kernel function for density estimation. The rotation kernel density can be fast estimated by a data structure named Rotation Kernel Density Sketch (RKDS). RKDS is a time- and memory- efficient method for kernel density estimation, even over data streams and distributed systems. RKDS is applicable for estimating density at specific points and also for representing data distribution. We provide theoretical analysis for rotation kernel and RKDS. Furthermore, we apply RKDS to outlier detection, concept drift detection, and personalized federated learning. Experiments show that our method improves time efficiency by up to 3 x 10(3) times compared with baselines. RKDS also provides comparable detecting precision and better delay on outlier detection and concept drift detection tasks.

关键词： Kernel Estimation Transforms Arrays Task analysis Anomaly detection Symbols Data stream sketch algorithm kernel density estimation

来源：评论

学校读者我要写书评

暂无评论

A Revisit to Graph Neighborhood Cardinality Estimation 40

A Revisit to Graph Neighborhood Cardinality Estimation

引用

40th IEEE International Conference on Data Engineering, ICDE 2024

作者： Wang, Pinghui Zhang, Yuanming Cheng, Kuankuan Zhao, Junzhou Xi'An Jiaotong University MOE Key Laboratory for Intelligent Networks and Network Security Xi'an China

ISBN: (纸本)9798350317152

Graph data are ubiquitous in real-world systems such as social networks and protein-protein interaction networks. In many applications, nodes usually are associated with real-value attributes, e.g., age, income, and wealth. Recently, industry and research communities have attracted attention to mining and learning attribute graphs. In this paper, we study the problem of calculating the general neighborhood cardinality of each node v in the graph, i.e., the sum of non-negative attribute values of the nodes in the k -hop neighborhood of a node v. The naive solution is to run a k -step breadth-first-search (BFS) algorithm starting from each node and storing all visited nodes' attributes. Clearly, the time complexity of this solution is Oleft(|V| dmaxk), where |V| is the number of nodes and dmax is the maximum node degree in the graph. In real-world networks such as Twitter, dmax is over 3×106. Therefore, it is infeasible to compute the neighborhood cardinality of nodes exactly in such massive networks even if we set k=2. To solve this problem, we propose efficient methods to compute the neighborhood cardinality of graphs with non-negative node attributes and binary node attributes, respectively. Extensive experiments on large real-world networks show the efficiency and effectiveness of our methods. © 2024 IEEE.

关键词： Proteins

来源：评论

学校读者我要写书评

暂无评论

A Multi-scale Perception Enhancement network for Underwater Waste Detection 9

A Multi-scale Perception Enhancement Network for Underwater ...

引用

9th International Conference on Image, Vision and Computing, ICIVC 2024

作者： Li, Yangke Zhang, Xinman MOE Key Lab for Intelligent Networks & Network Security School of Automation Science and Engineering Xi’an Jiaotong University Xi’an China

ISBN: (纸本)9798350385991

Underwater waste detection plays an important role in aquatic ecosystems and environmental protection. In fact, waste deposits in aquatic environments not only cause water contamination, but also pose potential risks to human health. However, underwater images often face the problem of image degradation, which makes underwater waste detection tasks more challenging. Therefore, we propose an efficient multi-scale perception enhancement network (MPE-Net) for underwater waste detection, which can provide more discriminative information to improve detection performance. On the one hand, we introduce a multi-scale semantic enhancement module (MSEM) to capture richer contextual features in different receptive fields, and enhance critical semantic information by using frequency domain information. On the other hand, we design a multi-scale boundary enhancement module (MBEM) to extract boundary information at different scales, and emphasize important detail features by using hybrid boundary perception. We conduct comprehensive experiments on two common underwater waste image datasets. Relevant experimental results show that our method outperforms other representative deep learning models. © 2024 IEEE.

关键词： Health risks

来源：评论

学校读者我要写书评

暂无评论

Simulating learning methodology(SLeM):an approach to machine learning automation

引用

National Science Review 2024年第8期11卷 12-17页

作者： Zongben Xu Jun Shu Deyu Meng School of Mathematics and Statistics Xi'an Jiaotong University Ministry of Education Key Lab of Intelligent Networks and Network Security Xi'an Jiaotong University Peng Cheng Laboratory Pazhou Laboratory(Huangpu)

MACHINE LEARNING REQUIRES AUTOMATION Machine learning （ML） is a fundamental technology of artificial intelligence（AI） that focuses on searching the possibly existing mapping f:x→y to fit a given dataset D={（xi,y... 详细信息

关键词： automation learning approach machine methodology simulating slem

来源：评论

学校读者我要写书评

暂无评论

intelligent management of waste bins in indoor public places based on waste detection and recognition

引用

COMPUTERS & ELECTRICAL ENGINEERING 2025年 123卷

作者： Li, Yangke Zhang, Xinman Xi An Jiao Tong Univ Fac Elect & Informat Engn Sch Automat Sci & Engn MOE Key Lab Intelligent Networks & Network Secur Xian Peoples R China

Due to the lack of convenient waste bins nearby, many people choose to litter indiscriminately in indoor public places. Dirty floors not only diminish the shopping experience of customers, but also increase potential risks to pedestrian safety. For the intelligent management of waste bins in indoor public places, this paper proposes a novel solution based on automatic waste detection and recognition, which helps to minimize littering and improve the recycling efficiency of indoor waste. This solution mainly uses a deep learning model to detect and recognize waste items, which can effectively record the distribution of waste and provide a basis for the reasonable placement of waste bins. Specifically, we construct a high-quality indoor waste image dataset for waste detection, which can provide an effective data source for model optimization of this task. This dataset is collected from three common public places, including hospitals, supermarkets, and subway stations. At the same time, it contains four waste categories, 4000 color images, and 6968 box-level annotations. In addition, we propose a novel feature decoupling network for indoor waste detection and recognition, which disentangles specific features for different vision tasks. On the one hand, it uses a recognition enhancement module to generate discriminative feature maps with more semantic information. On the other hand, it introduces a detection enhancement module to output rich feature maps with more detail information. As plug-and-play modules, these two modules are suitable for different networks. Relevant experimental results show that our method is competitive with other representative object detection models and can achieve consistent performance improvements on distinct models. In general, our solution provides new insights into indoor waste bin management.

关键词： Waste detection Waste recognition Waste bins Indoor public places Feature decoupling network

来源：评论

学校读者我要写书评

暂无评论

Hierarchical Task Planning for Power Line Flow Regulation

引用

CSEE Journal of Power and Energy Systems 2024年第1期10卷 29-40页

作者： Chenxi Wang Youtian Du Yanhao Huang Yuanlin Chang Zihao Guo the Ministry of Education Key Lab for Intelligent Networks and Network Security Xi'an Jiaotong UniversityXi'an 713599China the State Key Laboratory of Power Grid Safety and Energy Conservation China Electric Power Research InstituteBeijing 100192China

The complexity and uncertainty in power systems cause great challenges to controlling power *** a popular data-driven technique,deep reinforcement learning(DRL)attracts attention in the control of power ***,DRL has some inherent drawbacks in terms of data efficiency and *** paper presents a novel hierarchical task planning(HTP)approach,bridging planning and DRL,to the task of power line flow ***,we introduce a threelevel task hierarchy to model the task and model the sequence of task units on each level as a task planning-Markov decision processes(TP-MDPs).Second,we model the task as a sequential decision-making problem and introduce a higher planner and a lower planner in HTP to handle different levels of task *** addition,we introduce a two-layer knowledge graph that can update dynamically during the planning procedure to assist *** results conducted on the IEEE 118-bus and IEEE 300-bus systems demonstrate our HTP approach outperforms proximal policy optimization,a state-of-the-art deep reinforcement learning(DRL)approach,improving efficiency by 26.16%and 6.86%on both systems.

关键词： Knowledge graph power line flow regulation reinforcement learning task planning

来源：评论

学校读者我要写书评

暂无评论

Lightweight deep learning model for underwater waste segmentation based on sonar images

引用

WASTE MANAGEMENT 2024年 190卷 63-73页

作者： Li, Yangke Zhang, Xinman Xi An Jiao Tong Univ Fac Elect & Informat Engn Sch Automat Sci & Engn MOE Key Lab Intelligent Networks & Network Secur Xian 710049 Shaanxi Peoples R China

In recent years, the rapid accumulation of marine waste not only endangers the ecological environment but also causes seawater pollution. Traditional manual salvage methods often have low efficiency and pose safety risks to human operators, making automatic underwater waste recycling a mainstream approach. In this paper, we propose a lightweight multi-scale cross-level network for underwater waste segmentation based on sonar images that provides pixel-level location information and waste categories for autonomous underwater robots. In particular, we introduce hybrid perception and multi-scale attention modules to capture multi-scale contextual features and enhance high-level critical information, respectively. At the same time, we use sampling attention modules and cross-level interaction modules to achieve feature down-sampling and fuse detailed features and semantic features, respectively. Relevant experimental results indicate that our method outperforms other semantic segmentation models and achieves 74.66 % mIoU with only 0.68 M parameters. In particular, compared with the representative PIDNet Small model based on the convolutional neural network architecture, our method can improve the mIoU metric by 1.15 percentage points and can reduce model parameters by approximately 91 %. Compared with the representative SeaFormer T model based on the transformer architecture, our approach can improve the mIoU metric by 2.07 percentage points and can reduce model parameters by approximately 59 %. Our approach maintains a satisfactory balance between model parameters and segmentation performance. Our solution provides new insights into intelligent underwater waste recycling, which helps in promoting sustainable marine development.

关键词： Underwater waste Sonar images Semantic segmentation Lightweight model Multi-scale perception Cross-level interaction

来源：评论

学校读者我要写书评

暂无评论

Graph contrastive learning with high-order feature interactions and adversarial Wasserstein-distance-based alignment

引用

INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS 2025年第5-6期16卷 3449-3460页

作者： Wang, Chenxu Wan, Zhizhong Meng, Panpan Wang, Shihao Wang, Zhanggong Xi An Jiao Tong Univ Sch Software Engn Xian 710049 Peoples R China Xi An Jiao Tong Univ MoE Key Lab Intelligent Networks & Network Secur Xian 710049 Peoples R China

Graph contrastive learning (GCL) has proven to be an effective approach for unsupervised representation learning on graph-structured data. However, existing GCL models face two major limitations. First, existing feature augmentation methods fail to capture the high-order interactions among raw features, which are essential for feature engineering. Second, effective strategies for extracting global information from graphs for contrastive learning remain limited. To address these limitations, we propose a novel GCL model with high-order feature interactions and adversarial Wasserstein-distance-based alignment. Our model employs DNNs to capture complex interactions among raw features and introduce an alignment-based loss function to effectively extract global graph information. While traditional methods for calculating Wasserstein distance between graph views are computationally intensive, we overcome this challenge by training an adversarial Wasserstein-distance discriminator that enables efficient distance computation. We conduct extensive experiments on five benchmark datasets to evaluate the performance of the proposed method. The experimental results demonstrate that our approach achieves superior performance on classification tasks.

关键词： Graph contrastive learning Feature interaction Adversarial Wasserstein-distance-based alignment

来源：评论

学校读者我要写书评

暂无评论

Physically Realizable Adversarial Creating Attack Against Vision-Based BEV Space 3D Object Detection

引用

IEEE TRANSACTIONS ON IMAGE PROCESSING 2025年 34卷 538-551页

作者： Wang, Jian Li, Fan Lv, Song He, Lijun Shen, Chao Xi An Jiao Tong Univ Sch Informat & Commun Engn Shaanxi Key Lab Deep Space Explorat Intelligent In Xian 710049 Peoples R China Xi An Jiao Tong Univ Sch Cyber Sci & Engn MOE Key Lab Intelligent Networks & Network Secur Xian 710049 Peoples R China

Vision-based 3D object detection, a cost-effective alternative to LiDAR-based solutions, plays a crucial role in modern autonomous driving systems. Meanwhile, deep models have been proven susceptible to adversarial examples, and attacking detection models can lead to serious driving consequences. Most previous adversarial attacks targeted 2D detectors by placing the patch in a specific region within the object's bounding box in the image, allowing it to evade detection. However, attacking 3D detector is more difficult because the adversary may be observed from different viewpoints and distances, and there is a lack of effective methods to differentiably render the 3D space poster onto the image. In this paper, we propose a novel attack setting where a carefully crafted adversarial poster (looks like meaningless graffiti) is learned and pasted on the road surface, inducing the vision-based 3D detectors to perceive a non-existent object. We show that even a single 2D poster is sufficient to deceive the 3D detector with the desired attack effect, and the poster is universal, which is effective across various scenes, viewpoints, and distances. To generate the poster, an image-3D applying algorithm is devised to establish the pixel-wise mapping relationship between the image area and the 3D space poster so that the poster can be optimized through standard backpropagation. Moreover, a ground-truth masked optimization strategy is presented to effectively learn the poster without interference from scene objects. Extensive results including real-world experiments validate the effectiveness of our adversarial attack. The transferability and defense strategy are also investigated to comprehensively understand the proposed attack.

关键词： Three-dimensional printing Detectors Roads Object detection Autonomous vehicles Pipelines Feature extraction Visualization Solid modeling Perturbation methods Physical adversarial attack visual 3D detection universal patch adversarial robustness

来源：评论

学校读者我要写书评

暂无评论

CMW-Net: an adaptive robust algorithm for sample selection and label correction

引用

National Science Review 2023年第6期10卷 19-23页

作者： Jun Shu Xiang Yuan Deyu Meng School of Mathematics and Statistics Xi'an Jiaotong University Ministry of Education Key Lab of Intelligent Networks and Network Security Xi'an Jiaotong University Peng Cheng Laboratory Pazhou Laboratory (Huangpu)

PROBLEM Deep neural networks (DNNs) have recently achieved impressive performance in various *** successis largely attributed to massive but carefully labeled data expected to properly and sufficiently simulate testing ***,the data collection and annotation processes are often expensiveand time consuming,making obtained training datasets in real-life applications error prone,especially with data biasfrom ideal testing distributions.

关键词：

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：