检索结果-内蒙古大学图书馆

International Symposium on system and Software Reliability (ISSSR)

作者： Jinbo Zhao Jinling He Hu Song Xinyun Cheng Xiaolong Xu Jiangsu Key Laboratory of Big Data Security & Intelligent Processing Nanjing China State Grid Jiangsu Electric Power Company Limited Nanjing China School of Computer Science Nanjing University of Posts and Telecommunications Nanjing China

ISBN: (数字)9798350362930

ISBN: (纸本)9798350362947

The primary objective of outlier detection is to identify values that are significantly different from other data in the dataset. However, most of the current algorithms are effective for small-scale data and their performance is highly dependent on the choice of parameters. To this, the paper proposes an innovative outlier detection algorithm, named FBOD, which is based on data features. Specifically, FBOD utilizes the number of combinations of attribute values in each tuple across the dataset as a key feature of the tuple. This feature reflects the degree of differentiation of each data tuple from the overall dataset. By identifying data features that are significantly different from objects in the entire dataset, FBOD can accurately determine outliers. Based on the knowledge of attribute importance in Rough Set Theory, FBOD can efficiently reduce the data dimensionality, so high-dimensional datasets can be handled by FBOD. Furthermore, this paper employs a distributed parallelism approach in the algorithm. The performance of FBOD is evaluated on several datasets, including standard datasets and artificial datasets, and is compared with common outlier detection algorithms. Experimental results indicate that, compared to existing typical algorithms, FBOD achieves higher outlier detection accuracy without relying on a prior knowledge of datasets or parameter selection. It also proves the suitability of FBOD for large-scale dataset processing on distributed platforms.

关键词： Accuracy Distributed databases Rough sets Feature extraction Safety Object recognition security

来源：评论

学校读者我要写书评

暂无评论

BLRGCN: A dynamic traffic flow prediction model based on spatiotemporal graph convolutional network

BLRGCN: A dynamic traffic flow prediction model based on spa...

引用

IEEE International Conference on data Mining Workshops (ICDM Workshops)

作者： Qiuhao Shi Xiaolong Xu Xuanyan Liu Jiangsu Key Laboratory of Big Data Security & Intelligent Processing Nanjing University of Posts and Telecommunications Nanjing China School of Computer Science Nanjing University of Posts and Telecommunications Nanjing China

Because of the complex topology of urban road and the dynamic changes of traffic information over time, how to accurately and efficiently predict traffic flow has become a difficult problem. In order to capture spatiotemporal dependence at the same time, in this paper we propose BLRGCN, a dynamic traffic flow prediction model based on spatiotemporal graph convolutional network. This model improves the performance of bidirectional long short-term memory network (Bi-LSTM) by using residual network (ResNet), which can directly connect the input short to the output of nonlinear layer, and then combines the improved Bi-LSTM with graph convolutional neural network (GCN). BLRGCN model uses graphs to construct the network information of urban roads. In the graph, nodes express roads and edges express relationship between roads. The attributes of nodes describe traffic information on roads. We use GCN to extract the topological features of road network to capture spatial dependence. Besides, the BLRGCN model uses Bi-LSTM to obtain the information of the whole input sequence and extract the dynamic change characteristics of traffic data on the road to capture the temporal dependence. Compared with other methods on two traffic datasets, BLRGCN model has improved RMSE index by 0.53%-37.03% and 23.69%-43.83% respectively. The MAE index was improved by 34.89%-37.78% and 7.80%-51.88%, the accuracy index was improved by 0.19%-82.53% and 4.55%- 10.45%, the R2 index was improved by 0.24%-25.65% and 16.67%-29.94%, and the var index is improved by 0.25%-24.24% and 16.99%-30.39% respectively, which indicates that BLRGCN model has good prediction performance.

关键词：

来源：评论

学校读者我要写书评

暂无评论

ConvTrans-TPS: A Convolutional Transformer Model for Disk Failure Prediction in Large-Scale Network Storage systems

ConvTrans-TPS: A Convolutional Transformer Model for Disk Fa...

引用

International Conference on Computer Supported Cooperative Work in Design

作者： Shicheng Xu Xiaolong Xu Jiangsu Key Laboratory of Big Data Security & Intelligent Processing Nanjing University of Posts and Telecommunications Nanjing China School of Computer Science Nanjing University of Posts and Telecommunications Nanjing China

Disk failure is one of the most important reliability problems in large-scale network storage systems. Disk failure may lead to serious data loss and even disastrous consequences if the missing data cannot be recovered. Therefore, predicting disk failures is an important means of ensuring storage security in network storage systems. However, since the fault data in the fast degradation stage is smaller than the healthy data in the normal state, the mixture of healthy data and faulty data leads to extremely unbalanced data, which brings great challenges to finding hidden fault information, thus making fault prediction more accurate. Aiming at the above problems, a convolutional transformer model ConvTrans-TPS model for disk failure prediction in large-scale network storage systems is proposed. The ConvTrans-TPS model acquires dependencies between long-term sequence data through transformers and uses convolutional projections for attention computation to enhance attention to local contextual information. data augmentation to predict failures in the next 7 days. Validated by the analysis on the Backblaze dataset, the F1 is 0.96 and the Matthews correlation coefficient (MCC) is 0.92. Compared with the popular CNN-LSTM model in recent years, our proposed method improves F1 and MCC by 4% and 5%, respectively, improving the prediction accuracy.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Improving Code Search with Multi-Modal Momentum Contrastive Learning

Improving Code Search with Multi-Modal Momentum Contrastive ...

引用

IEEE/ACM International Conference on Program Comprehension (ICPC)

作者： Zejian Shi Yun Xiong Yao Zhang Zhijie Jiang Jinjing Zhao Lei Wang Shanshan Li Shanghai Key Laboratory of Data Science School of Computer Science Fudan University China Peng Cheng Laboratory Shenzhen China School of Computer National University of Defense Technology China National Key Laboratory of Science and Technology on Information System Security China

Contrastive learning has recently been applied to enhancing the BERT-based pre-trained models for code search. However, the existing end-to-end training mechanism cannot sufficiently utilize the pre-trained models due to the limitations on the number and variety of negative samples. In this paper, we propose MoCoCS, a multi-modal momentum contrastive learning method for code search, to improve the representations of query and code by constructing large-scale multi-modal negative samples. MoCoCS increases the number and the variety of negative samples through two optimizations: integrating multi-batch negative samples and constructing multi-modal negative samples. We first build momentum contrasts for query and code, which enables the construction of large-scale negative samples out of a mini-batch. Then, to incorporate multi-modal code information, we build multi-modal momentum contrasts by encoding the abstract syntax tree and the data flow graph with a momentum encoder. Experiments on CodeSearchNet with six programming languages demonstrate that our method can further improve the effectiveness of pre-trained models for code search.

关键词：

来源：评论

学校读者我要写书评

暂无评论

DEGSTalk: Decomposed Per-Embedding Gaussian Fields for Hair-Preserving Talking Face Synthesis

DEGSTalk: Decomposed Per-Embedding Gaussian Fields for Hair-...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Kaijun Deng Dezhi Zheng Jindong Xie Jinbao Wang Weicheng Xie Linlin Shen Siyang Song Computer Vision Institute School of Computer Science and Software Engineering Shenzhen University National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Guangdong Provincial Key Laboratory of Intelligent Information Processing Department of Computer Science University of Exeter

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Accurately synthesizing talking face videos and capturing fine facial features for individuals with long hair presents a significant challenge. To tackle these challenges in existing methods, we propose a decomposed per-embedding Gaussian fields (DEGSTalk), a 3D Gaussian Splatting (3DGS)-based talking face synthesis method for generating realistic talking faces with long hairs. Our DEGSTalk employs Deformable Pre-Embedding Gaussian Fields, which dynamically adjust pre-embedding Gaussian primitives using implicit expression coefficients. This enables precise capture of dynamic facial regions and subtle expressions. Additionally, we propose a Dynamic Hair-Preserving Portrait Rendering technique to enhance the realism of long hair motions in the synthesized videos. Results show that DEGSTalk achieves improved realism and synthesis quality compared to existing approaches, particularly in handling complex facial dynamics and hair preservation. Our code is available at https://***/CVI-SZU/DEGSTalk.

关键词： Hair Training Three-dimensional displays Dynamics Signal processing Rendering (computer graphics) Noise measurement Speech processing Faces Videos

来源：评论

学校读者我要写书评

暂无评论

Anomaly detection and traceback scheme for cloud-edge networks

Anomaly detection and traceback scheme for cloud-edge networ...

引用

International Conference on Computer Supported Cooperative Work in Design

作者： Xuanyan Liu Jinling He Hu Song Xinyun Cheng Luyun Hu Xiaolong Xu Jiangsu Key Laboratory of Big Data Security & Intelligent Processing Nanjing China State Grid Jiangsu Electric Power Company Limited Nanjing China School of Computer Science Nanjing University of Posts and Telecommunications Nanjing China

ISBN: (数字)9798350349184

ISBN: (纸本)9798350349191

Cloud-edge networks (CEN) has revolutionized the traditional network architecture. However, there are many security issues in CEN needing to be solved. In this case, CEN administrator should be able to detect network anomalies and find out the sources of anomalies rapidly, and respond in time to block anomalies. In order to protect network systems with CEN, we propose a mechanism consisting of two modules, including anomaly detection and anomaly traceback. The self-similarity anomaly detection is introduced for the first time to detect anomalies. In addition, we also propose an anomaly traceback method which utilizes the characteristics of CEN. An CEN testbed is established to evaluate the mechanism we proposed. Results of experiments show that the proposed mechanism can discover anomalies quickly and accurately, and trace anomaly sources and find anomaly paths accurately.

关键词： Federated learning Refining Process control Network security Network architecture Control systems Indexes

来源：评论

学校读者我要写书评

暂无评论

security Compressed Sensing Image Encryption Algorithm Based on Elliptic Curve 9th

Security Compressed Sensing Image Encryption Algorithm Based...

引用

9th International Conference of Pioneering Computer Scientists, Engineers and Educators, ICPCSEE 2023

作者： Jin, Anan Li, Xiang Xiong, Qingzhi School of Information Engineering East China University of Technology Jiangxi Nanchang330013 China Jiangxi Engineering Technology Research Center of Nuclear Geoscience Data Science and System East China University of Technology Jiangxi Nanchang330013 China Jiangxi Key Laboratory of Cybersecurity Intelligent Perception East China University of Technology Jiangxi Nanchang330013 China

ISBN: (纸本)9789819959679

Aiming to address the problems of high costs associated with the storage and transmission of environmental monitoring images, as well as potential security risks, this paper proposes a security compressed sensing image encryption algorithm based on elliptic curve cryptography. The algorithm introduces elliptic curve encryption technology within a compressed sensing framework, using elliptic curve cryptography to encrypt the matrix during the compression perception acquisition process. This enables secure acquisition of environmental monitoring encrypted images. Therefore, this paper presents a security compressed sensing framework to address security gaps in the compression perception reconstruction process and improve the security of environmental monitoring images. Experimental results show that the proposed algorithm effectively encrypts and reconstructs environmental monitoring images, with high security and resistance to cracking. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Compressed sensing

来源：评论

学校读者我要写书评

暂无评论

Lgccn: A Label-Guided Contrastive Capsule Network for Few-Shot Text Classification

SSRN

引用

SSRN 2023年

作者： Xu, Yang Xu, Xiaolong Xiao, Fu Jiangsu Key Laboratory of Big Data Security & Intelligent Processing Nanjing University of Posts and Telecommunications Nanjing China School of Computer Science Nanjing University of Posts and Telecommunications Nanjing China

Few-shot text classification aims to learn transferable knowledge from a limited dataset to perform classification tasks in unseen domains. Recently, metric-based meta-learning methods have demonstrated their advantages in few-shot text classification. However, the influence of randomly sampling support set on prototype quality has often been overlooked by these methods. In this paper, we propose a Label-Guided Contrastive Capsule Network (LGCCN) to address this issue. LGCCN leverages label information to guide supervised contrastive learning, using labels as anchors to enhance the clustering process, resulting in higher-quality sentence vectors. Furthermore, LGCCN employs label-guided dynamic routing process to iteratively generate a prototype, significantly mitigating the impact of random support set sampling. We conducted extensive experiments on four datasets, comparing LGCCN with recent methods. The results reveal that LGCCN, compared to the recent best work, has achieved an improvement of 10.1% and 2.1% in classification accuracy for 1-shot and 5-shot tasks, respectively. Additionally, ablation studies validate the effectiveness of each module in our approach. © 2023, The Authors. All rights reserved.

关键词： Iterative methods

来源：评论

学校读者我要写书评

暂无评论

End-to-End Remote Sensing Image Scene Classification with Vision Transformers 23

End-to-End Remote Sensing Image Scene Classification with Vi...

引用

6th International Conference on Big data Technologies, ICBDT 2023

作者： Yuan, Hao Liu, Kun Shi, Jiechuan Wang, Can Wang, Weiwei School of Information Science and Engineering University of Jinan Jinan250022 China China High-resolution Earth Observation System Shandong Center of Data and Application University of Jinan Jinan250022 China Shandong Provincial Key Laboratory of Network Based Intelligent Computing University of Jinan Jinan250022 China QiLu Research Institute Jinan250200 China

ISBN: (纸本)9798400707667

In recent years, the development of deep learning technology has led to widespread attention on Vision Transformer (ViT) as an emerging image classification method. Remote sensing image classification is an important task in the field of remote sensing, with extensive application prospects. This paper aims to explore the remote sensing image classification method based on Vision Transformer, addressing the limitations of traditional convolutional neural networks in terms of global perception capability, context information retrieval, and positional encoding. The classification performance of the Vision Transformer model is evaluated and compared on remote sensing datasets. Vision Transformer is a deep neural network model based on self-attention mechanism that can capture the global context information in images and has achieved remarkable performance in various computer vision tasks. Furthermore, experimental results demonstrate that the remote sensing image classification method based on Vision Transformer exhibits outstanding accuracy and generalization ability. Compared to traditional convolutional neural networks, it can better capture the global features in remote sensing images and has better scalability when dealing with large-scale remote sensing image data. Experimental results on different remote sensing image datasets show that the model performs well compared to state-of-the-art methods. Specifically, Vision Transformer achieves average classification accuracies of 95.41%, 98.26%, 93.74% and 95.25% on the AID, UC-Merced, NWPU-RESISC45 and Optimal31 datasets, respectively. © 2023 ACM.

关键词： Neural network models

来源：评论

学校读者我要写书评

暂无评论

Evaluation of Continual Learning Methods for Object Hardness Recognition

Evaluation of Continual Learning Methods for Object Hardness...

引用

2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023

作者： Liu, Yiwen Fang, Senlin Liu, Chengliang Wang, Jingnan Mai, Ke Zhang, Yupo Yi, Zhengkun Wu, Xinyu Shenzhen Institute of Advanced Technology Chinese Academy of Sciences Guangdong Provincial Key Laboratory of Robotics and Intelligent System Shenzhen518055 China University of Chinese Academy of Sciences Beijing100049 China Shenzhen Institute of Artificial Intelligence and Robotics for Society Siat Branch Shenzhen518055 China City University of Macau Faculty of Data Science 999078 China

ISBN: (纸本)9798350327182

How to solve the problem of continual learning of robotic tactile perception in an open, dynamic environment is a pressing task. Although continual learning has been extensively studied in computer vision, the evaluation of those algorithms does not provide insights on whether their solutions may help continual learning in the context of tactile robotics. In this paper, we propose a continual learning scenario in the robotic tactile domain, and design experiments with tactile hardness dataset that allow us to compare the performance of many continual learning algorithms in the tactile domain. We examine the performance of regularization-based and replay methods in the tactile domain and discuss the effect of replay number on model performance. We show that regularization-based methods do not perform well on the tactile dataset. On the contrary, replay methods can maintain better anti-forgetting performance, but the number of model parameters will be larger, and the training time will be longer. We also found that the replay ratio should be higher than 0.2 to maintain the performance of the replay method on the tactile hardness dataset. © 2023 IEEE.

关键词： Robotics

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：