检索结果-内蒙古大学图书馆

3rd International Symposium on computer Applications and Information Systems, ISCAIS 2024

作者： Zhuang, Hongwei Liu, Weisheng School of Computer and Software Engineering Liaoning University of Science and Technology Liaoning Anshan China

ISBN: (纸本)9781510681316

Underwater target detection is an important method for marine life detection. However, the accuracy of target detection and recognition is affected by the problems of image occlusion, blurred water quality and complex background of underwater targets. An improved YOLOV7 underwater biological target detection method is proposed to address the above problems. Firstly, the DCN structure is improved by incorporating the residual structure into the DCN, with new jump connections and 1×1 convolutional branches, so that the network can learn the residual information between the input and the output. It can improve the expressive ability of the model, reduce the number of parameters, and enhance the adaptability to the shape of the object. Finally, COT3 is integrated, which utilises a self-attention mechanism to weight the input feature maps so that the model can focus on important contextual information around the target and improve the model's ability to handle complex scenes. The experimental results show that the improved YOLOv7 model achieves an average detection accuracy mAP value of 80.4%, which is 2.8% higher than the original algorithm, and the algorithm is better than the original YOLOv7 algorithm in terms of detection accuracy and speed, which is more suitable for the detection of underwater targets. © 2024 SPIE.

关键词： Water quality

来源：评论

学校读者我要写书评

暂无评论

Uvcgan-Dehaze: a dehazing method for unpaired images

引用

Soft Computing 2024年第20期28卷 12217-12226页

作者： Li, Canlin Zhang, Xiangfei Zhang, Wenjiao Su, Haowen Bi, Lihua School of Computer Science and Technology Zhengzhou University of Light Industry Zhengzhou China School of Software Engineering Zhengzhou University of Light Industry Zhengzhou China

To solve the problems of feature loss and color difference after image dehazing and poor dehazing effect in real hazy images, a method UVCGAN-Dehaze is proposed for unpaired image dehazing. In the proposed model, the generator learns the relationship between image features and low-frequency features at multiple scales by combining the U-Net with the transformer. We strengthen the connection between spatial and channel features by adding Convolutional Block Attention Module(CBAM) before the skip connection of each layer of U-Net of the generator and adaptively adjusting the weight allocation of the feature map, so that the image after dehazing can better retain the detailed information of the image. Simultaneously, in order to reduce the feature loss and color difference after image dehazing, we propose a loss function named Feature Laplacian Perception (FLP) loss, which jointly constrains the generator by fusing the perception loss and Laplacian loss, so that the generated image reduces the feature loss and color difference. The method not only generates dehazing images with high perceptual quality in terms of important regions and details, but also avoids content and color loss. Extensive experiments verify the effectiveness of the method on synthetic images and real hazy images, and the proposed method achieves better dehazing results when compared with the experimental results of previous typical methods. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024.

关键词： Color

来源：评论

学校读者我要写书评

暂无评论

Music source separation via hybrid waveform and spectrogram based generative adversarial network

引用

Multimedia Tools and Applications 2024年 1-15页

作者： Wu, Qiuxia Deng, Haipeng Hu, Kun Wang, Zhiyong School of Software Engineering South China University of Technology Guangdong Province Guangzhou China School of Computer Science The University of Sydney SydneyNSW Australia

Music source separation aims to disentangle individual sources from the mixture of musical signals. Existing generative adversarial network (GAN) based methods generally work on the spectrogram domain only. However, this practice ignores the patterns from the waveform domain, which are more informative for modelling some categories of sources. In this paper, we propose a fully hybrid GAN framework to integrate knowledge from both domains. In particular, the generator formulates acoustical patterns from waveform and spectrogram domains, while the discriminator provides discriminative information based on the local patch-level spectrograms such that the generator can produce more plausible separation results. Furthermore, to enhance the quality of estimated sources, we devise a perceptual spectrogram loss term, which is a complement of the waveform-level loss. The proposed method is evaluated on two widely used music source separation datasets, producing music sources of high signal-to-distortion ratio (12.03 in MIR-1K dataset and 8.08 in MUSDB18 dataset). These results demonstrate the superiority of the proposed method compared with the state-of-the-art methods. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.

关键词： Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

CK-index: A Distribution-Aware Learned Index for Composite Keys 22

CK-index: A Distribution-Aware Learned Index for Composite K...

引用

22nd IEEE International Symposium on Parallel and Distributed Processing with Applications, ISPA 2024

作者： Wei, Zhengyang Ye, Baoliu Cai, Miao Hohai University College of Computer and Software China Nanjing University Department of Computer Science and Technology China Nanjing University of Aeronautics and Astronautics College of Computer Science and Technology China

ISBN: (纸本)9798331509712

The learned index is a high-performance index structure that uses machine learning methods to predict key positions in a large key space efficiently. Existing learned indexes suffer from underfitting of key-to-position mapping, leading to poor lookup performance. This paper finds that a data distribution property in the widely-used composite key schema addresses this issue effectively. Specifically, the composite key consists of an agglomerate of attributes. Keys with the same attribute value have a regular data distribution, which leads to a higher fitness of key-to-position mapping. Applying the property, we introduce CK-index, a distribution-aware learned index for composite keys. CK-index divides the key space according to attribute values and trains each learned model separately for an attribute to achieve high fitness of key-to-position mapping. Furthermore, it achieves low data storage consumption via storing composite key's attributes instead of the whole keys. We evaluate the CK-index using real-world datasets. Evaluation results demonstrate that CK-index performs much better in lookup performance, bulk loading time and space consumption compared to B+Tree, RMI, PGM-index and ALEX. © 2024 IEEE.

关键词： Composite key Data distribution Learned index

来源：评论

学校读者我要写书评

暂无评论

LQG Problem of Descriptor Systems and Separation Principle 63

LQG Problem of Descriptor Systems and Separation Principle

引用

63rd IEEE Conference on Decision and Control, CDC 2024

作者： Nosrati, Komeil Belikov, Juri Tepljakov, Aleksei Petlenkov, Eduard Tallinn University of Technology Department of Computer Systems Estonia Tallinn University of Technology Department of Software Science Estonia

ISBN: (纸本)9798350316339

The performance of linear-quadratic regulators (LQRs), which possess robustness with guaranteed levels of gain and phase margin, will be affected by incomplete state information. By furnishing this regulator with a linear-quadratic estimator (LQE), a recursive algorithm with an optimal feedback law can be obtained. Notwithstanding this, the optimal operation of this so-called linear-quadratic Gaussian (LQG) algorithm might be compromised in transformed stochastic descriptor systems if constraints are violated. Inspired by recent results in stochastic control, we deal with the LQG problem using a direct strategy without any transformations and regularity assumptions. First, an expected quadratic cost function using a regularized least-squares is formulated. Using the estimated states with a designed LQE, we then connect the formulation to a constrained recursive minimization problem under Bellman's principle. To accomplish this, a dynamic programming approach is used in a backward policy within a finite horizon to develop an LQG algorithm for the original system. We conclude the study by stating the separation principle of optimal control and state estimation and verifying the results, where despite noisy measurements, the controller effectively track the reference position while stabilizing the DC motor. © 2024 IEEE.

关键词： Cost functions

来源：评论

学校读者我要写书评

暂无评论

Self-Supervised Exploration via Temporal Inconsistency in Reinforcement Learning

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2024年第11期5卷 5530-5539页

作者： Gao, Zijian Xu, Kele Zhai, Yuanzhao Ding, Bo Feng, Dawei Mao, Xinjun Wang, Huaimin National University of Defense Technology School of Computer Changsha410000 China State Key Laboratory of Complex and Critical Software Environment Beijing100191 China

In sparse extrinsic reward settings, reinforcement learning remains a challenge despite increasing interest in this field. Existing approaches suggest that intrinsic rewards can alleviate issues caused by reward sparsity. However, many studies overlook the critical role of temporal information, essential for human curiosity. This article introduces a novel intrinsic reward mechanism inspired by human learning processes, where curiosity is evaluated by comparing current observations with historical knowledge. Our method involves training a self-supervised prediction model, periodically saving snapshots of the model parameters, and employing the nuclear norm to assess the temporal inconsistency between predictions from different snapshots as intrinsic rewards. Additionally, we propose a variational weighting mechanism to adaptively assign weights to the snapshots, enhancing the model's robustness and performance. Experimental results across various benchmark environments demonstrate the efficacy of our approach, which outperforms other state-of-the-art methods without incurring additional training costs and exhibits higher noise tolerance. Our findings indicate that leveraging temporal information in intrinsic rewards can significantly improve exploration performance, motivating future research to develop more robust and accurate reward systems for reinforcement learning. © 2024 IEEE.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Weighted Attribute Based Conditional Proxy Re-Encryption in the Cloud

引用

computers, Materials & Continua 2025年第4期83卷 1399-1414页

作者： Xixi Yan Jing Zhang Pengyu Cheng Henan Key Laboratory of Network Cryptography Technology Network Information Security and Cryptography LaboratorySchool of SoftwareHenan Polytechnic UniversityJiaozuo454003China School of Computer Science and Technology Henan Polytechnic UniversityJiaozuo454000China

Conditional proxy re-encryption(CPRE)is an effective cryptographic primitive language that enhances the access control mechanism and makes the delegation of decryption permissions more granular,but most of the attribute-based conditional proxy re-encryption(AB-CPRE)schemes proposed so far do not take into account the importance of user attributes.A weighted attribute-based conditional proxy re-encryption(WAB-CPRE)scheme is thus designed to provide more precise decryption rights *** introducing the concept of weight attributes,the quantity of system attributes managed by the server is reduced *** the same time,a weighted tree structure is constructed to simplify the expression of access structure *** conditional proxy re-encryption,large amounts of data and complex computations are outsourced to cloud servers,so the data owner(DO)can revoke the user’s decryption rights directly with minimal *** scheme proposed achieves security against chosen plaintext attacks(CPA).Experimental simulation results demonstrated that the decryption time is within 6–9 ms,and it has a significant reduction in communication and computation cost on the user side with better functionality compared to other related schemes,which enables users to access cloud data on devices with limited resources.

关键词： Cloud service conditional proxy re-encryption user revocation weighted attribute

来源：评论

学校读者我要写书评

暂无评论

An enterprise portrait tag extraction method based on context embedding and knowledge distillation

引用

Soft Computing 2024年第20期28卷 11823-11834页

作者： Li, Xiang Ding, Xingshuo Zhu, Quanyin Ma, Jialin Zhou, Hong Sun, Jizhou Faculty of Computer and Software Huaiyin Institute of Technology Jiangsu Huaian223003 China Qingdao Ocean Shipping Mariners College Shandong Qingdao266400 China

Enterprise portraits are reliant on the meticulous process of tag extraction, with the accuracy and efficiency of this foundational task being pivotal for the construction of high-quality portraits and their subsequent applicability in varied real-world scenarios. However, current methodologies, chiefly semantic clustering and deep neural networks, are constrained by significant shortcomings: the former struggles with intricate text feature discernment, while the latter, although high-performing, are marked by their extensive computational requirements. In response to these challenges, we present EPCEKD, a novel approach leveraging context embedding and knowledge distillation, designed to refine the tag extraction process. EPCEKD utilizes structured enterprise context data to facilitate feature expansion and impose tag constraints, enabling the efficient vectorization and integration of enterprise context data and text. Further, EPCEKD amalgamates a large-scale BERT network and a streamlined GRU network, embedding the branch network into the backbone network to forge a unified tag extraction model. To validate the efficacy of EPCEKD, comprehensive experiments were conducted on the Ente-pku dataset and benchmarked against eight prevailing models, with EPCEKD demonstrating superior performance, achieving 95.17% in precision, recall, and F1-score, significantly surpassing models like BiLSTM-CNN (94.25%) and Transformer (91.46%). Importantly, EPCEKD not only excels in precision and F1-score but also enhances the computing speed of large-scale neural networks by 4-11.4 times, showcasing its substantial potential for practical enterprise portrait applications. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

ON THE EFFECT OF BATCH SIZE IN BYZANTINE-ROBUST DISTRIBUTED LEARNING 12

ON THE EFFECT OF BATCH SIZE IN BYZANTINE-ROBUST DISTRIBUTED ...

引用

12th International Conference on Learning Representations, ICLR 2024

作者： Yang, Yi-Rui Shi, Chang-Wei Li, Wu-Jun National Key Laboratory for Novel Software Technology Department of Computer Science and Technology Nanjing University Nanjing China

Byzantine-robust distributed learning (BRDL), in which computing devices are likely to behave abnormally due to accidental failures or malicious attacks, has recently become a hot research topic. However, even in the independent and identically distributed (i.i.d.) case, existing BRDL methods will suffer a significant drop on model accuracy due to the large variance of stochastic gradients. Increasing batch size is a simple yet effective way to reduce the variance. However, when the total number of gradient computation is fixed, a too-large batch size will lead to a too-small iteration number (update number), which may also degrade the model accuracy. In view of this challenge, we mainly study the effect of batch size when the total number of gradient computation is fixed in this work. In particular, we show that when the total number of gradient computation is fixed, the optimal batch size corresponding to the tightest theoretical upper bound in BRDL increases with the fraction of Byzantine workers. Therefore, compared to the case without attacks, a larger batch size is preferred when under Byzantine attacks. Motivated by the theoretical finding, we propose a novel method called Byzantine-robust stochastic gradient descent with normalized momentum (ByzSGDnm) in order to further increase model accuracy in BRDL. We theoretically prove the convergence of ByzSGDnm for general non-convex cases under Byzantine attacks. Empirical results show that when under Byzantine attacks, using a relatively large batch size can significantly increase the model accuracy, which is consistent with our theoretical results. Moreover, ByzSGDnm can achieve higher model accuracy than existing BRDL methods when under deliberately crafted attacks. In addition, we empirically show that increasing batch size has the bonus of training acceleration. © 2024 12th International Conference on Learning Representations, ICLR 2024. All rights reserved.

关键词： Stochastic systems

来源：评论

学校读者我要写书评

暂无评论

Efficient Federated Learning Using Dynamic Update and Adaptive Pruning with Momentum on Shared Server Data

引用

ACM Transactions on Intelligent Systems and technology 2024年第6期15卷 1-28页

作者： Liu, Ji Jia, Juncheng Zhang, Hong Yun, Yuhui Wang, Leye Zhou, Yang Dai, Huaiyu Dou, Dejing Hithink RoyalFlush Information Network Co. Ltd. Hangzhou China School of Computer Science and Technology Soochow University Suzhou China Collaborative Innovation Center of Novel Software Technology and Industrialization Nanjing China Baidu Research Beijing China Key Lab of High Confidence Software Technologies Ministry of Education and Software Institute Peking University Beijing China Department of Computer Science and Software Engineering Auburn University AuburnAL United States Department of Electrical and Computer Engineering North Carolina State University RaleighNC United States Bedi Cloud and School of Computer Science Fudan University Beijing China

Despite achieving remarkable performance, Federated Learning (FL) encounters two important problems, i.e., low training efficiency and limited computational resources. In this article, we propose a new FL framework, i.e., FedDUMAP, with three original contributions, to leverage the shared insensitive data on the server in addition to the distributed data in edge devices so as to efficiently train a global model. First, we propose a simple dynamic server update algorithm, which takes advantage of the shared insensitive data on the server while dynamically adjusting the update steps on the server in order to speed up the convergence and improve the accuracy. Second, we propose an adaptive optimization method with the dynamic server update algorithm to exploit the global momentum on the server and each local device for superior accuracy. Third, we develop a layer-adaptive model pruning method to carry out specific pruning operations, which is adapted to the diverse features of each layer so as to attain an excellent tradeoff between effectiveness and efficiency. Our proposed FL model, FedDUMAP, combines the three original techniques and has a significantly better performance compared with baseline approaches in terms of efficiency (up to 16.9 times faster), accuracy (up to 20.4% higher), and computational cost (up to 62.6% smaller). © 2024 Copyright held by the owner/author(s). Publication rights licensed to ACM.

关键词： Federated learning

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：