检索结果-内蒙古大学图书馆

Self-knowledge distillation with dimensional history knowledge

science China(Information sciences) 2025年

作者： Wenke HUANG Mang YE Zekun SHI He LI Bo DU National Engineering Research Center for Multimedia Software School of Computer ScienceWuhan University

Existing self-knowledge distillation (Self-KD) solutions usually focus on transferring historical predictions of individual instances to the current network. However, this approach tends to create overconfidence for easy instances and underconfidence for hard instances. The widely used temperature-based strategies to smooth or sharpen the predicted distributions can lead to inconsistencies across instances, causing sensitivity issues. To address this, our approach views a queue of instances as an ensemble rather than treating each instance independently. We propose a novel method that distills historical knowledge from a dimensional perspective, utilizing intra class characteristics and interclass relationships within each ensemble. First, we align each dimension distribution from the current network to the historical output. Second, we ensure each dimension is closer to similar dimensions than dissimilar ones, maintaining consistent attitudes from present and historical perspectives. Our insights reveal that distilling historical knowledge from a dimensional perspective is more effective than the traditional instance-based approach, with potential applications in related tasks. Empirical results on three famous datasets and various network architectures demonstrate the superiority of our proposed method. Our code is available at https://***/WenkeHuang/DimSelfKD.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Mutual Retinex: Combining Transformer and CNN for Image Enhancement

引用

IEEE Transactions on Emerging Topics in Computational Intelligence 2024年第3期8卷 2240-2252页

作者： Jiang, Kui Wang, Qiong An, Zhaoyi Wang, Zheng Zhang, Cong Lin, Chia-Wen Harbin Institute of Technology School of Computer Science and Technology Harbin150001 China Wuhan University National Engineering Research Center for Multimedia Software School of Computer Science Wuhan430072 China Wuhan Polytechnic University School of Mathematics and Computer Science Wuhan430024 China National Tsing Hua University Department of Electrical Engineering Institute of Communications Engineering Hsinchu30013 Taiwan

Images captured in low-light or underwater environments are often accompanied by significant degradation, which can negatively impact the quality and performance of downstream tasks. While convolutional neural networks (CNNs) and Transformer architectures have made significant progress in computer vision tasks, there are few efforts to harmonize them into a more concise framework for enhancing such images. To this end, this study proposes to aggregate the individual capability of self-attention (SA) and CNNs for accurate perturbation removal while preserving background contents. Based on this, we carry forward a Retinex-based framework, dubbed as Mutual Retinex, where a two-branch structure is designed to characterize the specific knowledge of reflectance and illumination components while removing the perturbation. To maximize its potential, Mutual Retinex is equipped with a new mutual learning mechanism, involving an elaborately designed mutual representation module (MRM). In MRM, the complementary information between reflectance and illumination components are encoded and used to refine each other. Through the complementary learning via the mutual representation, the enhanced results generated by our model exhibit superior color consistency and naturalness. Extensive experiments have shown the significant superiority of our mutual learning based method over thirteen competitors on the low-light task and ten methods on the underwater image enhancement task. In particular, our proposed Mutual Retinex respectively surpasses the state-of-the-art method MIRNet-v2 by 0.90 dB and 2.46 dB in PSNR on the LOL 1000 and FIVEK datasets, while with only 19.8% model parameters. © 2017 IEEE.

关键词： Image enhancement

来源：评论

学校读者我要写书评

暂无评论

Blockchain-Assisted Fine-Grained Deduplication and Integrity Auditing for Outsourced Large-Scale Data in Cloud Storage

引用

IEEE Internet of Things Journal 2025年第12期12卷 21662-21678页

作者： Liu, Bingyun Zhang, Xiaojun Yang, Xingchun Zhang, Yuan Xue, Jingting Zhou, Rang Southwest Petroleum University School of Computer Science and Software Engineering Engineering Research Center for Intelligent Oil and Gas Exploration Chengdu610500 China Sichuan Police College Intelligent Policing Key Laboratory of SichuanProvince Luzhou646000 China University of Electronic Science and Technology of China School of Computer Science and Engineering Chengdu610054 China Chengdu University of Technology School of Computer Science and Cyber Security Chengdu610059 China

Cloud computing has emerged as a promising mode for storaging vast quantities of big data, which is vulnerable to potential security threats, making it urgent to ensure data confidentiality and integrity auditing. In addition, as a large number of duplicate data files exist in cloud storage, data deduplication is significant to improve storage efficiency. In this paper, we propose a blockchain-assisted fine-grained deduplication and integrity auditing scheme for outsourced large-scale data in cloud storage, achieving internal deduplication and cross-user external deduplication for ciphertexts and authentication tags. By constructing sparse summation ciphertext tree, the scheme implements proofs of ownership protocol through vector commitment, and guarantees the retrieval of distributed deduplication data blocks by designing the reconstruction matrix and bloom filter. The scheme exploits blockchain and smart contracts to ensure transparent integrity auditing without a third-party auditor (TPA), thereby avoiding malicious auditing biases. Security analysis and performance evaluation demonstrate the feasibility of the scheme for deploying in cloud storage systems. © 2014 IEEE.

关键词： Data integrity

来源：评论

学校读者我要写书评

暂无评论

Binary Program Vulnerability Mining Based on Neural Network

引用

computers, Materials & Continua 2024年第2期78卷 1861-1879页

作者： Zhenhui Li Shuangping Xing Lin Yu Huiping Li Fan Zhou Guangqiang Yin Xikai Tang Zhiguo Wang School of Information and Software Engineering University of Electronic Science and Technology of ChinaChengdu611731China School of Electrical and Computer Engineering University ofWaterlooWaterlooN2L 3G1Canada

software security analysts typically only have access to the executable program and cannot directly access the source code of the *** poses significant challenges to security *** it is crucial to identify vulnerabilities in such non-source code programs,there exists a limited set of generalized tools due to the low versatility of current vulnerability mining ***,these tools suffer from some *** terms of targeted fuzzing,the path searching for target points is not streamlined enough,and the completely random testing leads to an excessively large search ***,when it comes to code similarity analysis,there are issues with incomplete code feature extraction,which may result in information *** this paper,we propose a cross-platform and cross-architecture approach to exploit vulnerabilities using neural network obfuscation *** leveraging the Angr framework,a deobfuscation technique is introduced,along with the adoption of a VEX-IR-based intermediate language conversion *** combination allows for the unified handling of binary programs across various architectures,compilers,and compilation ***,binary programs are processed to extract multi-level spatial features using a combination of a skip-gram model with self-attention mechanism and a bidirectional Long Short-Term Memory(LSTM)***,the graph embedding network is utilized to evaluate the similarity of program *** on these similarity scores,a target function is determined,and symbolic execution is applied to solve the target *** solved content serves as the initial seed for targeted *** binary program is processed by using the de-obfuscation technique and intermediate language transformation method,and then the similarity of program functions is evaluated by using a graph embedding network,and symbolic execution is performed based on these similarity *** approach facilitates

关键词： Vulnerability mining de-obfuscation neural network graph embedding network symbolic execution

来源：评论

学校读者我要写书评

暂无评论

EmoNet: A Convolutional Neural Network and Transfer Learning-Based Framework for Facial Micro-Expression Recognition and Emoticonification 5

EmoNet: A Convolutional Neural Network and Transfer Learning...

引用

5th International Conference on Artificial Intelligence and Electromechanical Automation, AIEA 2024

作者： Gong, Rui Wang, Zhongyi Zhao, Jie Southwest Petroleum University School of Computer Science and Software Engineering Chengdu China

ISBN: (纸本)9798350366174

With the rising popularity of online social interactions, emojis play a pivotal role in communication, effectively conveying people's emotions. Hence, accurately converting facial micro-expressions into corresponding emojis becomes crucial. However, current micro-expression recognition technology faces challenges of low accuracy and poor generalization. To address these issues, this paper innovatively proposes a method that combines transfer learning with convolutional neural networks. This approach significantly enhances the model's generalization capability and recognition accuracy. Experimental results demonstrate notable improvements in both accuracy and robustness in micro-expression recognition, effectively overcoming common problems found in traditional methods, such as overfitting and sample imbalance. © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Temporal-Aware Deep Reinforcement Learning for Energy Storage Bidding in Energy and Contingency Reserve Markets

IEEE Transactions on Energy Markets, Policy and Regulation

引用

IEEE Transactions on Energy Markets, Policy and Regulation 2024年第3期2卷 392-406页

作者： Li, Jinhao Wang, Changlong Zhang, Yanru Wang, Hao Monash University Department of Data Science and AI Faculty of IT and Monash Energy Institute MelbourneVIC3800 Australia Monash University Department of Civil Engineering MelbourneVIC3800 Australia University of Electronic Science and Technology of China School of Computer Science and Engineering Chengdu611731 China

The battery energy storage system (BESS) has immense potential for enhancing grid reliability and security through its participation in the electricity market. BESS often seeks various revenue streams by taking part in multiple markets to unlock its full potential, but effective algorithms for joint-market participation under price uncertainties are insufficiently explored in the existing research. To bridge this gap, we develop a novel BESS joint bidding strategy that utilizes deep reinforcement learning (DRL) to bid in the spot and contingency frequency control ancillary services (FCAS) markets. Our approach leverages a transformer-based temporal feature extractor to effectively respond to price fluctuations in seven markets simultaneously and helps DRL learn the best BESS bidding strategy in joint-market participation. Additionally, unlike conventional 'black-box' DRL model, our approach is more interpretable and provides valuable insights into the temporal bidding behavior of BESS in the dynamic electricity market. We validate our method using realistic market prices from the Australian National Electricity Market. The results show that our strategy outperforms benchmarks, including both optimization-based and other DRL-based strategies, by substantial margins. Our findings further suggest that effective temporal-aware bidding can significantly increase profits in the spot and contingency FCAS markets compared to individual market participation. © 2023 IEEE.

关键词： Real time systems

来源：评论

学校读者我要写书评

暂无评论

Invisible Fingerprint-based Copyright Protection for Generated Medical Image 2

Invisible Fingerprint-based Copyright Protection for Generat...

引用

2nd IEEE International Conference on Control, Electronics and computer Technology, ICCECT 2024

作者： Yang, Nan Jia, Wenjun Deng, Peng Yang, Wen School of Computer Science and Software Engineering Southwest Petroleum University Nanchong China

ISBN: (纸本)9798350380958

In the recent years, the development of generative models has stimulated the health care progress, specially medical image generation. The synthetic medical images can be applied to several fields and have many utilization, such as data aug- mentation for model training. However, this might lead to the significant copyright issues. Firstly, the data for model training may be the privacy data belong to individuals. Secondly, owners of well-trained model may have burden computation cost. Both of these two stakeholders are not will to share model to others without any compensation or the risk of the image abuse. In order to protect the copyright of the generative model for medical image and reduce the risk of the image abuse, we propose Copyright Protection mechanism for Generated Medical Image (CP4GMI) by inducing invisible fingerprint. In other words, the generated medical image should have the unique fingerprint belong to the corresponding generative model, which can be examined when image abuse happens. Numerous experiments conducted on different medical image datasets shows that our approach can track the generative model in a high-accuracy level when the fingerprint in the generated image is mostly invisible. From the perspective of users and model inventors, we propose a copyright protection mechanism that introduces invisible fingerprints, which can be printed on the generated medical images, and can be used to protect the rights when the copyright is threatened. © 2024 IEEE.

关键词： Copyrights

来源：评论

学校读者我要写书评

暂无评论

Cooperation among Multiple Medical Institutions on Retinal Disease Identification Based on Federated Learning and Unlearning 4

Cooperation among Multiple Medical Institutions on Retinal D...

引用

4th Asia-Pacific Conference on Communications Technology and computer science, ACCTCS 2024

作者： Deng, Peng Jia, Wenjun Yang, Nan Xiao, Liwen School of Computer Science and Software Engineering Southwest Petroleum University Nanchong China

ISBN: (纸本)9798350359985

Due to the aging of the world's population, the incidence of retinal diseases is on the rise. Machine learning is expected to have a crucial role in identifying retinal disease. Multiple medical institutions cooperate on retinal disease identification using machine learning, which can effectively break down data silos, but it is often accompanied by privacy concerns. Federated learning (FL) solves the privacy problem in the process of collaborative model training, where data of each medical institution does not leave the local to avoid data leakage. With a series of privacy laws coming into place, it is critical to give medical institutions participating in FL training the right to be forgotten. Federated unlearning (FU) allows a medical institution requesting to exit the cooperation to eliminate its contribution from the global model. We simulate the scenarios of cooperation among multiple medical institutions on retinal disease identification, using data augmentation to deal with an imbalanced dataset and backdoor attacks to evaluate the effects of FU. Experimental results show that FL and FU have potential for application in cooperation among multiple medical institutions on retinal disease identification. © 2024 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Traffic sign detection algorithm based on improved YOLOv5 4

Traffic sign detection algorithm based on improved YOLOv5

引用

4th International Conference on computer Vision and Pattern Analysis, ICCPA 2024

作者： Ma, Chi Li, Qinrong Hu, Hui Li, Jingyan Guo, Qiang School of Computer Science and Engineering Huizhou University Huizhou516007 China School of Computer Science and Software Engineering University of Science and Technology Liaoning Anshan114051 China

ISBN: (纸本)9781510682528

This paper introduces an enhanced YOLOv5 algorithm tailored for real-world traffic sign detection applications. Through the incorporation of Coordinate Attention after the SPPF module of the YOLOv5 backbone, the YOLOv5 neck pays more attention to key areas in the image and preserves accurate positional information. In response to the fact that traffic signs are mostly small targets, the PAN and FPN in the original algorithm's neck network are upgraded by substituting BiFPN for the previous feature fusion method, improving the algorithm's ability to detect traffic signs of different sizes, specifically targeting improved detection accuracy for small-scale targets. To validate the effectiveness of these modifications, we compared our improved model with other object detection model on the TT100K dataset and conducted ablation experiments. The experiments result revealed that the enhanced algorithm achieved an mAP of 94.2%, surpassing the original YOLOv5 model by 6.9%. The detection speed was 49.2 FPS, meeting the real-time requirements. © 2024 SPIE.

关键词： Traffic signals

来源：评论

学校读者我要写书评

暂无评论

Secure Batch Deduplication without Dual Servers in Backup System

引用

IEEE Transactions on Dependable and Secure Computing 2024年 1-13页

作者： Zheng, Haoyu Zeng, Shengke Li, Hongwei Li, Zhijun Xihua College Xihua University Chengdu China school of Computer and Software Engineering Xihua University Chengdu China School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China MIG Group Cisco Systems Canada Co. Ottawa ON Canada

Cloud storage provides highly available and low cost resources to users. However, as massive amounts of outsourced data grow rapidly, an effective data deduplication scheme is necessary. This is a hot and challenging field, in which there are quite a few researches. However, most of previous works require dual-server fashion to be against brute-force attacks and do not support batch checking. It is not practicable for the massive data stored in the cloud. In this paper, we present a secure batch deduplication scheme for backup system. Besides, our scheme resists the brute-force attacks without the aid of other servers. The core idea of the batch deduplication is to separate users into different groups by using short hashes. Within each group, we leverage group key agreement and symmetric encryption to achieve secure batch checking and semantically secure storage. We also extensively evaluate its performance and overhead based on different datasets. We show that our scheme saves the data storage by up to 89.84%. These results show that our scheme is efficient and scalable for cloud backup system and can also ensure data confidentiality. IEEE

关键词： Semantics

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：