检索结果-内蒙古大学图书馆

IEEE International Conference on Big Data

作者： Hongyu Qiu Yongwei Wang Yonghui Xu Lizhen Cui Zhiqi Shen School of Computer Science and Engineering Nanyang Technological University Singapore SIAS of Zhejiang University & College of Computer Science Zhejiang University China School of Software & Joint SDU-NTU Centre for Artificial Intelligence Research (C-FAIR) Shandong University China

Data are invaluable in machine learning (ML), yet they raise significant privacy concerns. In the real world, data are often distributed across isolated silos, challenging conventional ML methods that centralize data. Federated learning (FL) offers a privacy-preserving solution that enables learning without direct data transfer. Meanwhile, the “right to be forgotten” sparks privacy-preserving methods from another viewpoint as machine unlearning, enabling data owners to erase specific data contributions from ML models. However, the invisibility of data in FL scenarios complicates effective local data removal, necessitating tailored unlearning algorithms for FL. Existing federated unlearning methods fall into approximate unlearning, leaving residual memorization of target data, consequently diminishing user trust. To bridge this gap, we propose FedCIO, a novel framework for exact federated unlearning, designed to efficiently manage precise data removal requests in FL scenarios. Specifically, the framework involves client clustering, isolation among clusters, and one-shot aggregation of cluster models. This framework facilitates efficient unlearning by retraining only a relevant model subset rather than from scratch. To enhance the capability to handle Non-Independent and Identically Distributed (Non-IID) data, we further introduce an advanced spectral clustering implementation based on model similarity for better cluster partitioning. Comprehensive evaluation across common FL datasets with varied distributions demonstrates the superior performance of our proposed framework.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Bidirectional Looking with A Novel Double Exponential Moving Average to Adaptive and Non-adaptive Momentum Optimizers

arXiv

引用

arXiv 2023年

作者： Chen, Yineng Li, Zuchao Zhang, Lefei Du, Bo Zhao, Hai National Engineering Research Center for Multimedia Software School of Computer Science Wuhan University Wuhan430072 China Hubei Luojia Laboratory Wuhan430072 China Department of Computer Science and Engineering Shanghai Jiao Tong University Shanghai200240 China

Optimizer is an essential component for the success of deep learning, which guides the neural network to update the parameters according to the loss on the training set. SGD and Adam are two classical and effective optimizers on which researchers have proposed many variants, such as SGDM and RAdam. In this paper, we innovatively combine the backward-looking and forward-looking aspects of the optimizer algorithm and propose a novel ADMETA (A Double exponential Moving averagE To Adaptive and non-adaptive momentum) optimizer framework. For backward-looking part, we propose a DEMA variant scheme, which is motivated by a metric in the stock market, to replace the common exponential moving average scheme. While in the forward-looking part, we present a dynamic lookahead strategy which asymptotically approaches a set value, maintaining its speed at early stage and high convergence performance at final stage. Based on this idea, we provide two optimizer implementations, ADMETAR and ADMETAS, the former based on RAdam and the latter based on SGDM. Through extensive experiments on diverse tasks, we find that the proposed ADMETA optimizer outperforms our base optimizers and shows advantages over recently proposed com- Copyright © 2023, The Authors. All rights reserved.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Partial label learning with dissimilarity propagation guided candidate label shrinkage 23

Partial label learning with dissimilarity propagation guided...

引用

Proceedings of the 37th International Conference on Neural Information Processing Systems

作者： Yuheng Jia Fuchao Yang Yongqiang Dong School of Computer Science and Engineering Southeast University Nanjing China and Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications (Southeast University) Ministry of Education China College of Software Engineering Southeast University Nanjing China School of Computer Science and Engineering Southeast University Nanjing China

In partial label learning (PLL), each sample is associated with a group of candidate labels, among which only one label is correct. The key of PLL is to disambiguate the candidate label set to find the ground-truth label. To this end, we first construct a constrained regression model to capture the confidence of the candidate labels, and multiply the label confidence matrix by its transpose to build a second-order similarity matrix, whose elements indicate the pairwise similarity relationships of samples globally. Then we develop a semantic dissimilarity matrix by considering the complement of the intersection of the candidate label set, and further propagate the initial dissimilarity relationships to the whole data set by leveraging the local geometric structure of samples. The similarity and dissimilarity matrices form an adversarial relationship, which is further utilized to shrink the solution space of the label confidence matrix and promote the dissimilarity matrix. We finally extend the proposed model to a kernel version to exploit the non-linear structure of samples and solve the proposed model by the inexact augmented Lagrange multiplier method. By exploiting the adversarial prior, the proposed method can significantly outperform state-of-the-art PLL algorithms when evaluated on 10 artificial and 7 real-world partial label data sets. We also prove the effectiveness of our method with some theoretical guarantees. The code is publicly available at https://***/Yangfc-ML/DPCLS.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Hcl-Classifier: Cnn and Lstm Based Hybrid Malware Classifier for Internet of Things(Iot)

SSRN

引用

SSRN 2022年

作者： Abdullah, Muhammed Amin Yu, Yongbin Adu, Kwabena Imrana, Yakubu Wang, Xiangxiang Jingye, Cai School of Information and Software Engineering University of Electronic Science and Technology of China Sichuan Province China School of Computer Science and Engineering University of Electronic Science and Technology of China China

This paper highlights a hybrid static classifier based on CNN and bi-directional LSTM for malware classification tasks in the IoT. Our approach learns and takes note of the nature and complex patterns of the Byte and Assemble files represented in one-dimensional images for better feature extraction. CNN is employed for automatic feature selection and extraction. Furthermore, extracted features are forwarded to the bi-directional LSTM for classification. Extensive experiments were performed using the Microsoft malware classification datasets. Experimental results show that our HCL-Classifier achieves 99.91%, thus 0.5% better than the state-of-the-artworks. Moreover, the least performed classifier among the base models, such as the random forest, used in this paper achieved an accuracy of 98.34%. We attribute this to the nature of our 1D image representation approach. This study also discovered that different files in the dataset contain specific features that differ from file to file, and this we proved visually and through experiment © 2022, The Authors. All rights reserved.

关键词： Internet of things

来源：评论

学校读者我要写书评

暂无评论

Can PoW Consensus Protocol Resist the Whale Attack?

Can PoW Consensus Protocol Resist the Whale Attack?

引用

IEEE International Conference on software Quality, Reliability and Security (QRS)

作者： Xueyong Sun Qihao Bao Bixin Li College of Software Engineering Southeast University Nanjing China School of Computer Science and Engineering Southeast University Nanjing China

Proof of Work (PoW) is the most widely used consensus protocol. However, due to the hash rate competition mechanism, longest chain principle, and transaction fee mechanism of the PoW consensus protocol, malicious nodes can launch attacks to obtain more relative revenue than honest mining, which will discourage honest miners from packing transactions into blocks and verifying blocks. As a result, the speed of the nodes reaching consensus in the network is slowed down, or even consensus cannot be reached, which ultimately affects the security of the PoW consensus *** this paper, the Markov Decision Process (MDP) is used to simulate the whale attack launched by malicious nodes, and evaluate the capability of PoW consensus protocol against the whale attack. The experimental results show that the PoW consensus protocol is secure in the Bitcoin network when the transaction fee is set in the range of 0.002-0.3 block rewards and the transaction volume should not exceed 21.09 block rewards. In addition, the PoW consensus protocol will be more secure with the adjustment of parameters such as the number of block confirmations, block generation interval and block size.

关键词： Whales Bitcoin software quality Resists Reinforcement learning Markov processes Proof of Work

来源：评论

学校读者我要写书评

暂无评论

Protein-Ligand Binding Affinity Prediction Using Deep Learning 18

Protein-Ligand Binding Affinity Prediction Using Deep Learni...

引用

18th International computer Conference on Wavelet Active Media Technology and Information Processing, ICCWAMTIP 2021

作者： Atwereboannah, Abena Achiaa Wu, Wei-Ping Ding, Lei Yussif, Sophyani B. Tenagyei, Edwin Kwadwo University of Electronic Science and Technology of China School of Computer Science and Engineering Chengdu China SipingSoft Co. Ltd. Tianfu Software Park Chengdu China

ISBN: (纸本)9781665413640

Protein-ligand prediction plays a key role in drug discovery. Nevertheless, many algorithms are over reliant on 3D structure representations of proteins and ligands which are often rare. Techniques that can leverage the sequence-level representations of proteins, ligands and pockets are thus required to predict binding affinity and facilitate the drug discovery process. We have proposed a deep learning model with an attention mechanism to predict protein-ligand binding affinity. Our model is able to make comparable achievements with state-of-the-art deep learning models used for protein-ligand binding affinity prediction. © 2021 IEEE.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

Diffusion Prior Interpolation for Flexibility Real-World Face Super-Resolution

arXiv

引用

arXiv 2024年

作者： Yang, Jiarui Dai, Tao Zhu, Yufei Li, Naiqi Li, Jinmin Xia, Shu-Tao College of Artificial Intelligence Nankai University Tianjin China Tsinghua Shenzhen International Graduate School Tsinghua University Shenzhen China College of Computer Science and Software Engineering Shenzhen University China

Diffusion models represent the state-of-the-art in generative modeling. Due to their high training costs, many works leverage pre-trained diffusion models’ powerful representations for downstream tasks, such as face super-resolution (FSR), through fine-tuning or prior-based methods. However, relying solely on priors without supervised training makes it challenging to meet the pixel-level accuracy requirements of discrimination task. Although prior-based methods can achieve high fidelity and high-quality results, ensuring consistency remains a significant challenge. In this paper, we propose a masking strategy with strong and weak constraints and iterative refinement for real-world FSR, termed Diffusion Prior Interpolation (DPI). We introduce conditions and constraints on consistency by masking different sampling stages based on the structural characteristics of the face. Furthermore, we propose a condition Corrector (CRT) to establish a reciprocal posterior sampling process, enhancing FSR performance by mutual refinement of conditions and samples. DPI can balance consistency and diversity and can be seamlessly integrated into pre-trained models. In extensive experiments conducted on synthetic and real datasets, along with consistency validation in face recognition, DPI demonstrates superiority over SOTA FSR methods. The code is available at https://***/JerryYann/DPI. © 2024, CC BY.

关键词： Interpolation

来源：评论

学校读者我要写书评

暂无评论

An Interpretable Mechanical Fault Diagnosis Approach Based on Dynamics Embedding Neural Network and Contrastive Learning for Label-Free Vibration Signals

SSRN

引用

SSRN 2023年

作者： Sun, Shilong Peng, Tengyi Zhou, Yu Zhang, Xiao School of Mechanical Engineering and Automation Harbin Institute of Technology Guangdong Shenzhen518055 China College of Computer Science and Software Engineering Shenzhen University Shenzhen China College of Computer Science South-Central Minzu University Hubei Wuhan430074 China

The industrial vibration signals usually have no labels, and the intelligent fault diagnosis has low interpretability for the results. The supervised learning algorithm, for example, deep neural networks, cannot be directly applied to distinguish different vibration faults. This study proposes an automatic fault diagnosis method based on the Bootstrap Your Own Latent contrastive learning and dynamical systems model discovery algorithm (BYOLDIS), which can deal with the unlabeled signals and explain the diagnosis results. Firstly, deriving the differential equations matrix to develop the faulty bearing dynamical systems. Secondly, investigating the contrastive learning network with a time-delay embedding matrix to reconstruct the coordinates of the fault-dynamical system. Finally, constructing a library of fault machine dynamic polynomial equations can offer physical model-based prior constraints to reduce the dependence of neural networks on data. Therefore, the above three steps can make BYOLDIS achieve effective label-free and explainable mechanical fault automatic diagnosis. The simulated and experimental studies are conducted to verify the validity and robustness of the proposed method. The case studies demonstrate that BYOLDIS can achieve an accurate bearing fault diagnosis and provide a dynamic explanation for the diagnostic results. Therefore, the proposed BYOLDIS is a promising diagnostic method for processing unlabeled vibrational data. © 2023, The Authors. All rights reserved.

关键词： Failure analysis

来源：评论

学校读者我要写书评

暂无评论

User Information Perception in Virtual Reality Environment

User Information Perception in Virtual Reality Environment

引用

Symposia and Workshops on Ubiquitous, Autonomic and Trusted Computing, UIC-ATC

作者： Enyao Chang Xiaoping Che Jingzhi Cui Chenxin Qu School of Software Engineering Beijing Jiaotong University Beijing China School of Computer Science Beihang University Beijing China

With the development of virtual reality (VR) technology, panoramic video, a new method that is the fusion of VR technology and panoramic video technology, have gradually emerged and developed rapidly. Nowadays, VR panoramic video is becoming the most important application of virtual reality technology. Although the amount of video information is large, the user acceptance rate is relatively low and the perception is weak, which has a certain negative impact on the popularity of VR. Currently, there are few types of research on the VR information perception field, and the existing researches lack a great division of user characteristics and panoramic video types. Therefore, this paper mainly discusses the following two questions: Is the user’s information perception level in the VR environment significantly better than that in the traditional media environment? Do event types and user characteristics in VR videos affect users’ perception of information? In response to the questions, a total of 20 participants were recruited for the research. Through the analysis of the statistical calculation, this paper draws the conclusion: Media type has a significant impact on user information perception and the perception in the VR environment is significantly better than that of traditional media. Besides, this paper also finds a phenomenon: the user characteristics and the proportion of event types in the video have an impact on the user information perception effect. It shows that the higher the level of education, the better the information perception effect (ages between 20-30); the higher the proportion of mobile events and emergencies in the video, the better the user information perception effect.

关键词：

来源：评论

学校读者我要写书评

暂无评论

M3BUNet: Mobile Mean Max UNet for Pancreas Segmentation on CT-Scans

arXiv

引用

arXiv 2024年

作者： Juwita, Juwita Hassan, Ghulam Mubashar Akhtar, Naveed Datta, Amitava Department of Computer Science and Software Engineering The University of Western Australia Australia School of Computing and Information Systems The University of Melbourne Australia Department of Informatics Syiah Kuala University Banda Aceh Indonesia

Segmenting organs in CT scan images is a necessary process for multiple downstream medical image analysis tasks. Currently, manual CT scan segmentation by radiologists is prevalent, especially for organs like the pancreas, which requires a high level of domain expertise for reliable segmentation due to factors like small organ size, occlusion, and varying shapes. When resorting to automated pancreas segmentation, these factors translate to limited reliable labeled data to train effective segmentation models. Consequently, the performance of contemporary pancreas segmentation models is still not within acceptable ranges. To improve that, we propose M3BUNet, a fusion of MobileNet and U-Net neural networks, equipped with a novel Mean-Max (MM) attention that operates in two stages to gradually segment pancreas CT images from coarse to fine with mask guidance for object detection. This approach empowers the network to surpass segmentation performance achieved by similar network architectures and achieve results that are on par with complex state-of-the-art methods, all while maintaining a low parameter count. Additionally, we introduce external contour segmentation as a preprocessing step for the coarse stage to assist in the segmentation process through image standardization. For the fine segmentation stage, we found that applying a wavelet decomposition filter to create multi-input images enhances pancreas segmentation performance. We extensively evaluate our approach on the widely known NIH pancreas dataset and MSD pancreas dataset. Our approach demonstrates a considerable performance improvement, achieving an average Dice Similarity Coefficient (DSC) value of up to 89.53%±1.82 and an Intersection Over Union (IOU) score of up to 81.16±0.03% for the NIH pancreas dataset, and 88.60%±1.48 DSC and 79.90%±2.19 IOU for the MSD Pancreas dataset. © 2024, CC BY-NC-SA.

关键词： computerized tomography

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：