检索结果-内蒙古大学图书馆

2023 China Automation Congress, CAC 2023

作者： Liao, Liang Wan, Liang Liu, Mingsheng Li, Shusheng College of Computer Science and Technology Guizhou University State Key Laboratory of Public Big Data Guiyang China School of Computer Science and Engineering Northeastern Univiversity Shenyang China

ISBN: (纸本)9798350303759

When some application scenarios need to use semantic segmentation technology, like automatic driving, the primary concern comes to real-time performance rather than extremely high segmentation accuracy. To achieve a good trade-off between speed and accuracy, two-branch architecture has been proposed in recent years. It treats spatial information and semantics information separately which allows the module to be composed of two networks both not heavy. However, the process of fusing features with two different scales becomes a performance bottleneck for many nowaday two-branch models. In this research, we design a new fusion mechanism for two-branch architecture which is guided by attention computation. To be precise, we use the Dual-Guided Attention (DGA) module we proposed to replace some multi-scale transformations with the calculation of attention which means we only use several attention layers of near linear complexity to achieve performance comparable to frequently-used multi-layer fusion. To ensure that our module can be effective, we use Residual U-blocks (RSU) to build one of the two branches in our networks which aims to obtain better multi-scale features. Extensive experiments on Cityscapes and CamVid dataset show the effectiveness of our method. On Cityscapes, our light version network without pretrain weight can achieve 71.1% mIoU at 163 FPS on a single Nvidia RTX 3070 using full resolution images(1024×2048pix). And the large version can achieve 77.9% mIoU with a speed of 43 FPS which still reaches the real-time criterion. Our code and module has been open sourced at https://***/LikeLidoA/Mymodule. © 2023 IEEE.

关键词： computer vision

来源：评论

学校读者我要写书评

暂无评论

DPC: Large Model Alignment Method based on Decoding Probability Correction

DPC: Large Model Alignment Method based on Decoding Probabil...

引用

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Huang, Yanyi Liu, Yuying Tian, Lei Zhang, Yue Zhao, Xuechen Zhou, Bin Ma, Guodong School of Computer National University of Defense Technology Changsha China School of Data and Computer Science Shandong Women's University Jinan China

Large language models (LLMs) demonstrate significant generative capabilities but often face ethical alignment and robustness challenges. Conventional alignment methods rely on extensive human-annotated data and require retraining, leading to high computational costs and resource demands. Therefore, we propose a novel approach, Decoding Probability Correction (DPC), that aligns frozen LLMs without additional training or annotated data. DPC dynamically adjusts the probability distribution during inference, ensuring the generated content aligns with human values in real-time. Additionally, DPC incorporates a discriminator-based backtracking mechanism, further enhancing content safety by re-evaluating and refining generation choices. Experimental results on datasets such as the HH and AdvBench show that DPC significantly reduces harmful outputs while maintaining high levels of informativeness and helpfulness. The proposed method offers a cost-effective and efficient solution for enhancing the ethical alignment of LLMs in real-world applications. © 2025 IEEE.

关键词： Normal distribution

来源：评论

学校读者我要写书评

暂无评论

Few-Shot Partial Multi-View Learning

引用

IEEE Transactions on Pattern Analysis and Machine Intelligence 2023年第10期45卷 11824-11841页

作者： Zhou, Yuan Guo, Yanrong Hao, Shijie Hong, Richang Luo, Jiebo Ministry of Education Hefei University of Technology The Key Laboratory of Knowledge Engineering with Big Data Hefei230009 China Hefei University of Technology The School of Computer Science and Information Engineering Hefei230009 China University of Rochester The Department of Computer Science RochesterNY14627 United States

It is often the case that data are with multiple views in real-world applications. Fully exploring the information of each view is significant for making data more representative. However, due to various limitations and failures in data collection and pre-processing, it is inevitable for real data to suffer from view missing and data scarcity. The coexistence of these two issues makes it more challenging to achieve the pattern classification task. Currently, to our best knowledge, few appropriate methods can well-handle these two issues simultaneously. Aiming to draw more attention from the community to this challenge, we propose a new task in this paper, called few-shot partial multi-view learning, which focuses on overcoming the negative impact of the view-missing issue in the low-data regime. The challenges of this task are twofold: (i) it is difficult to overcome the impact of data scarcity under the interference of missing views;(ii) the limited number of data exacerbates information scarcity, thus making it harder to address the view-missing issue in turn. To address these challenges, we propose a new unified Gaussian dense-anchoring method. The unified dense anchors are learned for the limited partial multi-view data, thereby anchoring them into a unified dense representation space where the influence of data scarcity and view missing can be alleviated. We conduct extensive experiments to evaluate our method. The results on Cub-googlenet-doc2vec, Handwritten, Caltech102, Scene15, Animal, ORL, tieredImagenet, and Birds-200-2011 datasets validate its effectiveness. The codes will be released at https://***/zhouyuan888888/UGDA. © 1979-2012 IEEE.

关键词： Neural networks

来源：评论

学校读者我要写书评

暂无评论

An Empirical Investigation of Docker Sockets for Privilege Escalation and Defensive Strategies 5

An Empirical Investigation of Docker Sockets for Privilege E...

引用

5th International Conference on Innovative data Communication Technologies and Application, ICIDCA 2024

作者： Rajyashree, R. Mathi, Senthilkumar Saravanan, G. Sakthivel, M. Department of Computer Science and Engineering Amrita School of Computing Amrita Vishwa Vidyapeetham Coimbatore India Department of Artificial Intelligence and Data Science Erode Sengunthar Engineering College Erode India Department of Computer Science and Engineering Erode Sengunthar Engineering College Erode India

Cloud-based infrastructures often leverage virtualization, but its implementation can be expensive. Traditional coding methods can lead to issues when transitioning code from one computing environment to another. In response, the container paradigm emerged to offer cost-effective and agile delivery. Containers differ from full machine virtualization by compactly encapsulating the entire software and its dependencies. Leveraging containers, developers can create more secure and efficient applications. Docker, a prominent containerization platform, facilitates the execution of docker images. The Docker Hub serves as a popular repository for various images. Given the importance of application security, especially in the face of threats like malware, ransomware, and data breaches, ensuring robust security is imperative. The paper investigates vulnerabilities within Docker containers and proposes defensive strategies to mitigate potential breaches. In addition, it investigates attacks involving Docker sockets and suggests preventive measures for non-root users. © 2024 The Authors. Published by Elsevier B.V.

关键词： Containers

来源：评论

学校读者我要写书评

暂无评论

Tensor Multi-Task Learning for Multi-View Representation of 3D Shape

Tensor Multi-Task Learning for Multi-View Representation of ...

引用

2023 China Automation Congress, CAC 2023

作者： Zhang, Yan Fu, Sujia Qu, Yanyun School of Mathematical Sciences Guizhou Normal University Guiyang China School of Big Data and Computer Science Guizhou Normal University Guiyang China Xiamen University Computer Science Department Xiamen China

ISBN: (纸本)9798350303759

Research on 3D representation has sprung up recently. However, most existing view-based methods lack interpretability and few researchers explore to mine high-order correlation for multi-view features aggregate in 3D shape representation. This paper presents a tensor multi-task model learning, T-MTL, which transforms the original multi-view classification problem into constructing a common space in which the multi-view representation can be compared. The proposed method can construct a tensor structure in the classifier weight subspace with multiple viewpoints and use a tensor low-rank constraint and Bregman discrepancy to capture high-order correlations between multi-view images and 3D models. In the classifier weight subspace, our method makes the inter-class divergence larger and the intra-class divergence smaller. Moreover, the proposed model can improve recognition performance further by unifying multiple visual features in a flexible manner. Abundant experiments on these challenging benchmark datasets show that our method achieves comparable results to the deep learning models approach. © 2023 IEEE.

关键词： Tensors

来源：评论

学校读者我要写书评

暂无评论

Real-Time Pothole Detection at night to prevent road accidents using YOLOV5 6

Real-Time Pothole Detection at night to prevent road acciden...

引用

6th International Conference on Recent Trends in Advance Computing, ICRTAC 2023

作者： Mandal, Gouranga De, Anurag Dey, Arindam Gangarapu, Sudeepthi Chekuru, Sharmila Rani Nandini, Chekuri Vellore Institute of Technology School of Computer Science and Engineering Andhra Pradesh 522237 India Vellore Institute of Technology School of Computer Science and Engineering with Specialization in Data Analytics Andhra Pradesh 522237 India

ISBN: (纸本)9798350394702

This article proposes a technique to create a pothole detection system that can operate both during the day and at night. This model was developed using the real-time object identification YOLO (You Only Look Once) technique. The potholes are located using a pre-trained model with YOLO v5. In its creation, the PyTorch framework is utilized. The first YOLO model created using the PyTorch framework is YOLOv5, which is a lot more user-friendly and lightweight. In a manner similar to real-time object detection systems, the proposed method displays potholes in real time and emphasizes them with boundary boxes for accurate identification. The accuracy achieved with this method is 97% during the day and 92% at night which is found to be satisfactory. © 2023 IEEE.

关键词： Image acquisition

来源：评论

学校读者我要写书评

暂无评论

Self-training through Classifier Disagreement for Cross-Domain Opinion Target Extraction 23

Self-training through Classifier Disagreement for Cross-Doma...

引用

2023 World Wide Web Conference, WWW 2023

作者： Sun, Kai Zhang, Richong Samuel, Mensah Nikolaos, Aletras Mao, Yongyi Liu, Xudong Beijing Advanced Innovation Center for Big Data and Brain Computing School of Computer Science and Engineering Beihang University Beijing China Department of Computer Science University of Sheffield Sheffield United Kingdom School of Electrical Engineering and Computer Science University of Ottawa Ottawa Canada

ISBN: (纸本)9781450394161

Opinion target extraction (OTE) or aspect extraction (AE) is a fundamental task in opinion mining that aims to extract the targets (or aspects) on which opinions have been expressed. Recent work focus on cross-domain OTE, which is typically encountered in real-world scenarios, where the testing and training distributions differ. Most methods use domain adversarial neural networks that aim to reduce the domain gap between the labelled source and unlabelled target domains to improve target domain performance. However, this approach only aligns feature distributions and does not account for class-wise feature alignment, leading to suboptimal results. Semi-supervised learning (SSL) has been explored as a solution, but is limited by the quality of pseudo-labels generated by the model. Inspired by the theoretical foundations in domain adaptation [2], we propose a new SSL approach that opts for selecting target samples whose model output from a domain-specific teacher and student network disagree on the unlabelled target data, in an effort to boost the target domain performance. Extensive experiments on benchmark cross-domain OTE datasets show that this approach is effective and performs consistently well in settings with large domain shifts. © 2023 ACM.

关键词： Sentiment analysis

来源：评论

学校读者我要写书评

暂无评论

FedNorm: Optimised Federated Learning Strategy Using Hybrid Regularisation Norm 12th

FedNorm: Optimised Federated Learning Strategy Using Hybrid ...

引用

12th International Conference on Big data and Artificial Intelligence, BDA 2024

作者： John, Binu Menon, Varun Adhikari, Mainak Department of Computer Science and Engineering SCMS School of Engineering and Technology Kerala Ernakulam683576 India APJ Abdul Kalam Technological University Kerala Thiruvananthapuram695016 India School of Data Science Indian Institute of Science Education and Research Vithura Kerala Thiruvananthapuram695551 India

ISBN: (纸本)9783031818202

Federated Learning (FL) has emerged as a game-changing method for training machine learning (ML) models on decentralized data sources without centralizing the data itself. Conventional FL algorithms frequently exhibit performance instability and inefficiency, especially when working with non-Independent and Identically Distributed (non-IID) data distributions. To overcome these difficulties, we present FedNorm, a new FL method that uses hybrid regularisation approach to improve model stability and accuracy by balancing sparsity and smoothness, to increase generalization across varied data distributions. Our experimental setup uses the MNIST dataset to train models in both Independent and Identically Distributed (IID) and non-IID situations. The findings show that FedNorm routinely outperforms classic algorithms like FedAvg, FedSGD, FedNova and FedSAM, with minimal test loss and high accuracy. FedNorm has an exceptional accuracy rating of 0.98 on both IID and non-IID data. FedNorm addresses the constraints of conventional FL algorithms, resulting in a strong and dependable solution for FL in diverse situations. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Achieving reliable and anti-collusive outsourcing computation and verification based on blockchain in 5G-enabled IoT

引用

Digital Communications and Networks 2022年第5期8卷 644-653页

作者： Linjie Wang Youliang Tian Jinbo Xiong School of Data Science Tongren UniversityTongren554300China State Key Laboratory of Public Big Data College of Computer Science and Technology Guizhou UniversityGuiyang550025China Fujian Provincial Key Laboratory of Network Security and Cryptology College of Computer and Cyber SecurityFujian Normal UniversityFuzhou350117China

Widespread applications of 5G technology have prompted the outsourcing of computation dominated by the Internet of Things(IoT)cloud to improve transmission efficiency,which has created a novel paradigm for improving the speed of common connected objects in ***,although it makes it easier for ubiquitous resource-constrained equipment that outsources computing tasks to achieve high-speed transmission services,security concerns,such as a lack of reliability and collusion attacks,still exist in the outsourcing *** this paper,we propose a reliable,anti-collusion outsourcing computation and verification protocol,which uses distributed storage solutions in response to the issue of centralized storage,leverages homomorphic encryption to deal with outsourcing computation and ensures data ***,we embed outsourcing computation results and a novel polynomial factorization algorithm into the smart contract of Ethereum,which not only enables the verification of the outsourcing result without a trusted third party but also resists collusion *** results of the theoretical analysis and experimental performance evaluation demonstrate that the proposed protocol is secure,reliable,and more effective compared with state-of-the-art approaches.

关键词： Outsourcing computation Ethereum Smart contract Public verification Homomorphic encryption

来源：评论

学校读者我要写书评

暂无评论

Understanding News Background Information Seeking from Readers' Perspective 25

Understanding News Background Information Seeking from Reade...

引用

25th IEEE International Conference on computer Supported Cooperative Work in Design, CSCWD 2022

作者： Feng, Zhenni Ding, Xianghua Jiang, Yanqi Wu, Yiying Zhang, Peng Fudan University School of Computer Science Shanghai Key Laboratory of Data Science Shanghai China University of Glasgow School of Computer Science United Kingdom

ISBN: (纸本)9781665405270

Background information is very important for news reading and understanding, especially with today's rapid new media development for news production and consumption. When researchers actively develop tools to automatically provide back-ground information for people, we still have little understanding of how people, from readers' perspective, engage with news and what background information is needed. To address such a gap, we conducted a study with a news reading platform, observed participants' news reading behaviors using the platform and interviewed them over the process. Our study highlighted that news reading is a learning activity between leisure and study, and their background information searching behavior is part of Gestalt knowing - to make their existing knowledge more complete, by deepening their prior understandings, broadening their horizons, filling the gaps, clearing their confusion, as well as tracking the progress of an event as a whole, so a more holistic understanding could be achieved. We end by discussing implications of our findings. © 2022 IEEE.

关键词： Leisure

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：