检索结果-内蒙古大学图书馆

Multi-scale network for single image deblurring based on ensemble learning module

Multimedia Tools and Applications 2025年第11期84卷 9045-9064页

作者： Wu, WenBo Pan, Yun Su, Na Wang, JingTao Wu, ShaoChuan Xu, ZeKun Yu, YouJian Liu, YaPeng State Key Laboratory of Media Convergence and Communication Communication University of China Beijing100024 China School of Computer and Cyberspace Security Communication University of China Beijing100024 China School of Electronics and Information Engineering Harbin Institute of Technology Harbin150080 China College of Computer and Control Engineering Northeast Forestry University Harbin150040 China School of Computer and Information Engineering Tianjin Chengjian University Tianjin300384 China School of Computer Science and Technology Anhui University Hefei230601 China

In this paper, we have identified two primary issues with current multi-scale image deblurring methods. On the one hand, the blurring scale is ignored. On the other hand, the context information of images is not fully utilized, and the spatial detail information of images is lost. To better solve the above problems, on top of that, this paper proposes a multi-scale network for single image deblurring. The designed network contains two parts: 1) Ensemble Learning Model: The Weak Learning Model in this module connects multiple weak learners in series to increase network depth. Moreover, through multi-scale feature ensemble module, the output features and intermediate features of different weak learner are integrated, and finally strong learner is used to enhance the features with semantic information at different scales;2) Spatial Feature Enhancement Module: This module can preserve spatial detail information of images and compensate for the disadvantages of losing image spatial details due to continuous downsampling in the Ensemble Learning Model structure. Our proposed deblurring model has been extensively evaluated on several benchmark datasets, and the model achieves superior performance compared to the most advanced deblurring methods. In particular, the peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) values are 33.57 dB and 0.965 on the GoPro dataset, respectively, with PSNR and SSIM improving by 0.49 dB and 0.002 over NAFNet, and 0.65 dB and 0.004 over Restormer. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.

关键词： Learning systems

来源：评论

学校读者我要写书评

暂无评论

A Deep Knowledge Tracing Model Based on Multi-Head Self-Attention Mechanism

A Deep Knowledge Tracing Model Based on Multi-Head Self-Atte...

引用

2023 International Conference on Algorithms, High Performance Computing, and Artificial Intelligence, AHPCAI 2023

作者： Zhou, Ya Wu, Fengzhen Huang, Guimin Lin, Jianxing Deng, Nanxiao Guo, Qingkai School of Computer Science and Information Security Guangxi Key Laboratory of Image and Graphic Intelligent Processing Guilin University of Electronic Technology Guilin541004 China

ISBN: (纸本)9781510671881

How to provide individualized instruction and personalized teaching for students and effectively track their knowledge status in a targeted manner is a key concern in the field of education. The target of knowledge tracing (KT) is to use the learner's behavior in historical learning activities to model the students' educational process and forcast the future performance of students. Recently, self-attention has been used several times in the KT field. Traditional models have the problem of losing information under long sequence data. Inspired by Transformer, In this paper, we present a deep knowledge tracing model (MA-DKT) based on the multi-head self-attention mechanism, which employs the LSTM model combined with the multi-head self-attention mechanism to implement a deep knowledge tracing model that is able to catch the data characteristics of the students at different temporal scales. Our model works well on publicly available datasets and achieves some improvement in AUC. © 2023 SPIE. All rights reserved.

关键词： Students

来源：评论

学校读者我要写书评

暂无评论

Optimization algorithm for global k-means algorithm 13

Optimization algorithm for global k-means algorithm

引用

13th International Conference on Intelligent Control and information Processing, ICICIP 2025

作者： Li, Sen Li, Fang Luo, Xiaonan School of Computer Science and Information Security Guilin University of Electronic Technology Guilin541004 China Guilin541004 China Guangxi Key Laboratory of Image and Graphic Intelligent Processing Guilin541004 China

ISBN: (纸本)9798331516147

The k-means method is widely utilized for clustering. Its simplicity, efficacy, and swiftness make it a favored choice among clustering algorithms. It faces the challenge of sensitivity to the initial class center. The enhancement strategy for the k-means algorithm focuses on refining the initial selection of cluster centroids, no longer using its random selection of initial class center points, but starting from the entire dataset, choosing the most distant point to serve as the starting cluster centroid. This can also prevent the local optimization problem of the k-means algorithm. The improved algorithm proposed in this article is an effective method that can select the initial class center point from a global perspective and reduce computation time. The suggested approach has undergone rigorous testing and comparative analysis across multiple datasets, yielding commendable outcomes. © 2025 IEEE.

关键词： Optimization algorithms

来源：评论

学校读者我要写书评

暂无评论

An Interference Vector Judgement Based User Clustering Scheme in Ultra Dense network 3

An Interference Vector Judgement Based User Clustering Schem...

引用

3rd International Conference on Advances in Electrical, Computing, Communication and Sustainable Technologies, ICAECT 2023

作者： Liang, Yanxia Zhao, Zhiheng Liu, Xin Zhao, Meng Zhao, Songlin Xi'An University of Posts and Telecommunications Dept. Shaanxi Key Laboratory of Information Communication Network and Security Xi'an China Xi'An Eurasia University Dept. School of Information Engineering Xi'an China

ISBN: (纸本)9781665494007

Ultra Dense network will undoubtedly play an important role in next generation communication network due to its excellent system performance. However, the deployment of large-scale small cells can cause serious interference problems among users. Therefore, this paper proposes an interference vector judgement based user clustering scheme to suppress the system throughput degradation problem caused by interference. For our proposed scheme, it first gives a design of inter-user interference weight to obtain user interference vectors;secondly, users are clustered by adjudicating the relative interference values between user interference vectors. In simulation, it can demonstrate that our proposed algorithm achieves significant improvements in enhancing system throughput, especially for cell-edge users, by comparing it with some other existing algorithms. © 2023 IEEE.

关键词： Vectors

来源：评论

学校读者我要写书评

暂无评论

Design and Implementation of an Automated APK Analysis System for Practical Forensics 13th

Design and Implementation of an Automated APK Analysis Syste...

引用

13th International Conference on computer Engineering and networks, CENet 2023

作者： Li, Yicheng Liang, Guangjun Tang, keyan Teng, Yijia Ruan, Haoran Mo, Zixi Department of Computer Information and Cyber Security Jiangsu Police Institute Nanjing China Engineering Research Center of Electronic Data Forensics Analysis Nanjing China Key Laboratory of Digital Forensics Department of Public Security of Jiangsu Province Nanjing China

ISBN: (纸本)9789819992386

In order to cope with the widespread emergence of malicious APK in the application market and the increasingly rampant situation of mobile malware, as well as to meet the needs of the actual forensics work, this paper, based on the analysis of Android security mechanisms and their security problems, analyses and summarises the methodological principles, main technologies and functional flaws of existing APK automated analysis tools, and proposes an APK automated analysis system for actual forensics. Compared with traditional systems, the system is easy to operate and has a higher recognition accuracy rate. Besides, it will perform machine learning on the analysis results and calculate the similarity through the model to effectively assist the forensic work and fill the demand for APK analysis in the field of combat. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd 2024.

关键词： Malware

来源：评论

学校读者我要写书评

暂无评论

Improving MPCitH with Preprocessing: Mask Is All You Need

IACR Transactions on Cryptographic Hardware and Embedded Sys...

引用

IACR Transactions on Cryptographic Hardware and Embedded Systems 2025年第2期2025卷 1-22页

作者： Liu, Guowei Liu, Guoxiao Jiang, Kaijie Yu, Qingyuan Jia, Keting Wei, Puwen Wang, Meiqin School of Cyber Science and Technology Shandong University Qingdao China Institute for Network Sciences and Cyberspace Tsinghua University Beijing China Institute for Advanced Study Tsinghua University Beijing China Quan Cheng Shandong Laboratory Jinan China Zhongguancun Laboratory Beijing China BNRist Tsinghua University Beijing China Key Laboratory of Cryptologic Technology and Information Security Ministry of Education Shandong University Jinan China

The MPC-in-the-head with preprocessing (MPCitH-PP) paradigm presents a novel approach for constructing post-quantum digital signatures like Picnic3. This paper revisits the MPCitH-PP construction, analyzing both its offline and online phases and proposing a reformulation of the protocol. By identifying redundant computations in these phases, we optimize them into a single phase, thereby enhancing the efficiency of MPCitH-PP. Furthermore, we explore the independence of the mask, demonstrating that it can be calculated in parallel, which also enables the optimization of the masked witness calculation. Our optimized implementation of Picnic3 shows significant improvements. At the L1 security level, the optimal software implementation reduces MPCitH-PP calculation time to about 30% of the previous implementation. The optimal signature implementation costs about 78% of the previous implementation time. At the L5 security level, MPCitH-PP with parallelism optimal is reduced to about 26% of the previous solution’s time, and the optimal signature implementation runs at about 53% of the previous solution’s time. For the hardware implementation, our optimizations reduce the clock cycles of MPCitH-PP from r sequential rounds to a single parallel round, where r denotes the number of rounds in the LowMC algorithm, with little change in hardware usage, and perform better in AT product, especially for parallel computing. © 2025, Ruhr-University of Bochum. All rights reserved.

关键词： Masks

来源：评论

学校读者我要写书评

暂无评论

FLAP: Federated Learning Aggregation Scheme Based on Privileged Secret Sharing

FLAP: Federated Learning Aggregation Scheme Based on Privile...

引用

2023 International Conference on networking and network Applications, NaNA 2023

作者： Mu, Xianyu Tian, Youliang Xiong, Jinbo Wang, Shuai Gong, Boxiang College of Computer Science and Technology Guizhou University State Key Laboratory of Public Big Data Guiyang550025 China College of Computer and Cyber Security Fujian Normal University Fujian Provincial Key Laboratory of Network Security and Cryptology Fuzhou350117 China Guizhou Xishan Technology Co. Ltd China

ISBN: (纸本)9798350327380

Federated Learning (FL) is a privacy-preserving machine learning technique that trains models on client devices and only uploads new model gradients to servers for aggregation. However, transmitting the true gradients introduces the risk of reverse inference attack, which can compromise client privacy. To address this issue, we present the federated learning aggregation scheme based on privileged secret sharing (FLAP). Firstly, we present privileged secret sharing by combining the one-time pad with shamir secret sharing. Secondly, we introduce game theory to generate games between servers to resist collusion attacks. Finally, the analysis demonstrates that FLAP enables the server to successfully update the global gradient while ensuring the utmost protection of client privacy and FLAP achiev the effect of resisting collusion attack and reverse inference attack. We have proved through experiments that FLAP is 26.42% more efficient than the current scheme, and has strong robustness to drop out. © 2023 IEEE.

关键词： Game theory

来源：评论

学校读者我要写书评

暂无评论

Image inpainting via Smooth Tucker decomposition and Low-rank Hankel constraint

引用

International Journal of computers and Applications 2023年第6期45卷 421-432页

作者： Cai, Jing Jiang, Jiawei Wang, Yibing Zheng, Jianwei Xu, Honghui Department of Forensic Science Zhejiang Police College Hangzhou China Key Laboratory of Public Security Information Application Based on Big-Data Architecture Hangzhou China College of Computer Science and Technology Zhejiang University of Technology Hangzhou China

Image inpainting, aiming at exactly recovering missing pixels from partially observed entries, is typically an ill-posed problem. As a powerful constraint, low-rank priors have been widely applied in image inpainting to transform such problems into well-posed ones. However, the low-rank assumption of original visual data is only in an approximate mode, which in turn results in suboptimal recovery of fine-grained details, particularly when the missing rate is extremely high. Moreover, a single prior cannot faithfully capture the complex texture structure of an image. In this paper, we propose a joint usage of Smooth Tucker decomposition and Low-rank Hankel constraint (STLH) for image inpainting, which enables simultaneous capturing of the global low-rankness and local piecewise smoothness. Specifically, based on the Hankelization operation, the original image is mapped to a high-order structure for capturing more spatial and spectral information. By employing Tucker decomposition for optimizing the Hankel tensor and simultaneously applying Discrete Total Variation (DTV) to the Tucker factors, sharper edges are generated and better isotropic properties are enhanced. Moreover, to approximate the essential rank of the Tucker decomposition and avoid facing the uncertainty problem of the upper-rank limit, a reverse strategy is adopted to approximate the true rank of the Tucker decomposition. Finally, the overall image inpainting model is optimized by the well-known alternate least squares (ALS) algorithm. Extensive experiments show that the proposed method achieves state-of-the-art performance both quantitatively and qualitatively. Particularly, in the extreme case with 99% pixels missed, the results from STLH are averagely ahead of others at least 0.9dB in terms of PSNR values. © 2023 Informa UK Limited, trading as Taylor & Francis Group.

关键词： Pixels

来源：评论

学校读者我要写书评

暂无评论

Learning Label Shift Correction for Test-Agnostic Long-Tailed Recognition 41

Learning Label Shift Correction for Test-Agnostic Long-Taile...

引用

41st International Conference on Machine Learning, ICML 2024

作者： Wei, Tong Mao, Zhen Zhou, Zi-Hao Wan, Yuanyu Zhang, Min-Ling School of Computer Science and Engineering Southeast University Nanjing China Key Laboratory of Computer Network and Information Integration Southeast University Ministry of Education China School of Software Technology Zhejiang University Ningbo China

Long-tail learning primarily focuses on mitigating the label distribution shift between long-tailed training data and uniformly distributed test data. However, in real-world applications, we often encounter a more intricate challenge where the test label distribution is agnostic. To address this problem, we first theoretically establish the substantial potential for reducing the generalization error if we can precisely estimate the test label distribution. Motivated by the theoretical insight, we introduce a simple yet effective solution called label shift correction (LSC). LSC estimates the test label distribution within the proposed framework of generalized black box shift estimation, and adjusts the predictions from a pre-trained model to align with the test distribution. Theoretical analyses confirm that accurate estimation of test label distribution can effectively reduce the generalization error. Extensive experimental results demonstrate that our method significantly outperforms previous state-of-the-art approaches, especially when confronted with non-uniform test label distribution. Notably, the proposed method is general and complements existing long-tail learning approaches, consistently improving their performance. The source code is available at https://***/Stomach-ache/label-shift-correction. Copyright 2024 by the author(s)

关键词： Black-box testing

来源：评论

学校读者我要写书评

暂无评论

A Deep Reinforcement Learning Approach for Online Taxi Dispatching

A Deep Reinforcement Learning Approach for Online Taxi Dispa...

引用

2023 International Conference on networking and network Applications, NaNA 2023

作者： Wang, YuBo Zhang, TaoYi Wei, ZhiCheng Key Lab of Network and Information Security China College of Computer and Cyber Security China Hebei Prov. Eng. Res. Center for Supply Chain Big Data Analytics Security Hebei Normal University Hebei Province Shijiazhuang China

ISBN: (纸本)9798350327380

With the development of smart city transportation systems, developing reasonable dispatching strategies for idle ride-hailing vehicles has become an urgent research problem. In this paper, we address the short-sightedness issue of existing dispatching strategies and design a ride-hailing dispatching model based on deep reinforcement learning in the Markov decision process(MDP) framework to obtain long-term dispatching strategies with foresight. Firstly, we define the basic framework of agent and optimize the dispatching strategy of ride-hailing services by defining a metric that combines instant revenue with the demand situation of the next state in the action space as a reward function. Then, we construct an order-matching system simulator based on historical order data sets to facilitate interaction between the intelligent agent and the environment. Finally, based on the Bellman optimality equation, we approximate the value function using the Eval Net and Target Net, and train the model using Temporal Difference Learning. Simulation experiments demonstrate that this approach can increase driver income, provide more services to passengers, and effectively improve the operating efficiency of the transportation network. © 2023 IEEE.

关键词： Reinforcement 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：