检索结果-内蒙古大学图书馆

41st International Conference on Machine Learning, ICML 2024

作者： He, Yue Li, Dongbai Tian, Pengfei Yu, Han Liu, Jiashuo Zou, Hao Cui, Peng Department of Computer Science and Technology Tsinghua University Beijing China Qiuzhen College Tsinghua University Beijing China Zhongguancun Laboratory Beijing China

Despite notable progress in enhancing the capability of machine learning against distribution shifts, training data quality remains a bottleneck for cross-distribution generalization. Recently, from a data-centric perspective, there have been considerable efforts to improve model performance through refining the preparation of training data. Inspired by realistic scenarios, this paper addresses a practical requirement of acquiring training samples from various domains on a limited budget to facilitate model generalization to target test domain with distribution shift. Our empirical evidence indicates that the advance in data acquisition can significantly benefit the model performance on shifted data. Additionally, by leveraging unlabeled test domain data, we introduce a Domain-wise Active Acquisition framework. This framework iteratively optimizes the data acquisition strategy as training samples are accumulated, theoretically ensuring the effective approximation of test distribution. Extensive real-world experiments demonstrate our proposal's advantages in machine learning applications. The code is available at https://***/dongbaili/DAA. Copyright 2024 by the author(s)

关键词： Data assimilation

来源：评论

学校读者我要写书评

暂无评论

Deep transformer and few‐shot learning for hyperspectral image classification

引用

CAAI Transactions on Intelligence Technology 2023年第4期8卷 1323-1336页

作者： Qiong Ran Yonghao Zhou Danfeng Hong Meiqiao Bi Li Ni Xuan Li Muhammad Ahmad College of Information Science and Technology Beijing University of Chemical TechnologyBeijingChina Key Laboratory of Computational Optical Imaging Technology Aerospace Information Research InstituteChinese Academy of SciencesBeijingChina China Greatwall Technology Group Co. LtdShenzhenChina Department of Computer Science National University of Computer and Emerging SciencesChiniotPakistan

Recently,deep learning has achieved considerable results in the hyperspectral image(HSI)***,most available deep networks require ample and authentic samples to better train the models,which is expensive and inefficient in practical *** few‐shot learning(FSL)methods generally ignore the potential relationships between non‐local spatial samples that would better represent the underlying features of *** solve the above issues,a novel deep transformer and few‐shot learning(DTFSL)classification framework is proposed,attempting to realize fine‐grained classification of HSI with only a few‐shot ***,the spatial attention and spectral query modules are introduced to overcome the constraint of the convolution kernel and consider the information between long‐distance location(non‐local)samples to reduce the uncertainty of ***,the network is trained with episodes and task‐based learning strategies to learn a metric space,which can continuously enhance its modelling ***,the developed approach combines the advantages of domain adaptation to reduce the variation in inter‐domain distribution and realize distribution *** three publicly available HSI data,extensive experiments have indicated that the proposed DT‐FSL yields better results concerning state‐of‐the‐art algorithms.

关键词： deep learning feature extraction hyperspectral image classification

来源：评论

学校读者我要写书评

暂无评论

BiDect: Leveraging cascaded Bidirectional Long Short-Term Memory for enhanced contextual modeling in plant disease classification

引用

Journal of Agricultural, Biological and Environmental Statistics 2025年 1-23页

作者： Medjadba, Yasmine Drid, Hamza Bardou, Dalal Zhan, Ying LAMIE Laboratory Computer science department Batna 2 University Batna Algeria Computer Science and Math department Abbes Laghrour University khenchela Algeria School of Computer and Software Nanyang Institute of Technology Nanyang China

Plant diseases pose a serious threat to global food security and economic stability due to significant crop losses. Timely detection and management are essential to reduce these losses. Unlike manual methods, which are slow and error-prone, deep learning (DL) provides accurate, automated solutions that save both time and effort. This study presents BiDect, a Patch Context Modeling approach using cascaded Bidirectional Long Short-term memory (Bi-LSTM) for plant disease detection and classification. Images are first divided into smaller patches to capture local fea-tures that may be missed in global analysis. Individual Bi-LSTM networks process these patch features to capture spatial dependencies. The final Bi-LSTM layer combines these local insights to identify global patterns, enhancing classification accuracy. This cascaded approach focuses on processing smaller patches, enabling the model to capture intricate patterns effectively, which is particularly beneficial for large-scale datasets. We evaluated the method on diverse datasets, testing its performance in both real-world fields and controlled laboratory conditions. Comparison with state-of-the-art models, such as ResNet, DenseNet, InceptionV3, Xception, and VGG19, demonstrated that the cascaded BiLSTM network, combined with CNN features, outperforms these models, achieving superior classification accuracy for real-world applications.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Assessing the Impact Of Time Intervals On PM2.5 Prediction Using LSTM Neural Networks 3

Assessing the Impact Of Time Intervals On PM2.5 Prediction U...

引用

3rd International Conference on Advanced Electrical Engineering, ICAEE 2024

作者： Boulkamh, Selma Zeghib, Nadia Yaddaden, Yacine University of Constantine 2-Abdelhamid Mehri Lire Laboratory Constantine Algeria University of Quebec at Rimouski Department of Mathematics Computer Science and Engineering Canada

ISBN: (纸本)9798350349351

Air pollution is a global issue with profound implications to human health and environmental sustainability. Especially, PM2.5, which refers to particulate matter with a size of 2.5 microns or smaller, poses significant health risks. Therefore, accurately predicting PM2.5 levels is crucial. In this aim, various machine-learning methods are employed for PM2.5 prediction. The problem arises from the use of collected datasets at varied time intervals (5 minutes, 15 minutes, 1 hour, etc.), resulting in inconsistency across predictive approaches. While shorter time intervals seems to offer heightened sensitivity to changes, implying simpler prediction, this presumption may not be universally applicable. This paper investigates the impact of time intervals on PM2.5 prediction through a comparative analysis of datasets collected at different intervals (15 minutes, 1 hour, and 8 hours) using Long Short Term Memory Neural Networks (LSTM). By leveraging statistical analysis and predictive modeling techniques, we assess the dataset's distributions and evaluate prediction model performance. This study underscores the importance of interval selection in PM2.5 prediction and informs future efforts to enhance air quality surveillance and public health protection measures. © 2024 IEEE.

关键词： Prediction models

来源：评论

学校读者我要写书评

暂无评论

Enhancing Financial Accessibility: A Tailored UPI Payment Application for Divyangjan 10

Enhancing Financial Accessibility: A Tailored UPI Payment Ap...

引用

10th International Conference on Advanced Computing and Communication Systems, ICACCS 2024

作者： Bhowmik, Biswajit Sudhama, Kruthika K. Dongala, Joshitha R. Antony, Reshma T. Girish, K.K. National Institute of Technology Karnataka Maharshi Sushrut Cas Lab Brics Laboratory Department of Computer Science and Engineering Mangalore Surathkal575025 India

ISBN: (纸本)9798350384369

The emergence of financial technology (FinTech) has transformed the financial sector, introducing a new era characterized by state-of-the-art technologies that enhance speed, affordability, and accessibility. The proliferation of the internet and smartphones has further accelerated this transformation, fostering greater connectivity and global interaction. Subsequently, these advancements have significantly expanded financial inclusion, ensuring access to financial services for previously under-served populations. While the rise of FinTech has propelled financial inclusion for many, individuals with disabilities have not experienced commensurate improvements in their financial accessibility. As the banking sector increasingly migrates to online platforms, people with disabilities encounter barriers stemming from inaccessible websites, mobile applications, and online banking services. This paper introduces a specialized UPI payment application designed explicitly for individuals with disabilities. The objective is to integrate this underserved demographic into the digital financial landscape, fostering financial inclusion and enhancing access to essential financial services. © 2024 IEEE.

关键词： Decentralized finance

来源：评论

学校读者我要写书评

暂无评论

A Novel Deep Learning Framework for Contraband Items Recognition in Smart City Applications 2nd

A Novel Deep Learning Framework for Contraband Items Recogni...

引用

2nd International Conference on Sustainability: Developments and Innovations, ICSDI 2024

作者： Khalid, Shehzad Ahmed, Zubair Noor, Ayesha Sedik, Ahmed Asif, Waqar Department of Computer Engineering Bahria University Islamabad Pakistan Smart Systems Engineering Laboratory College of Engineering Prince Sultan University Riyadh Saudi Arabia Computer Science and Cyber Security University of West London London United Kingdom

ISBN: (纸本)9789819783441

Ensuring safety and security is paramount in today’s complex environment, and the effective detection of contraband items plays a pivotal role in achieving this objective. Contraband items, ranging from illegal substances to unauthorized goods, pose a threat to public safety, security, and the overall well-being of smart city inhabitants. Such items are currently detected by human operator reviewing the images from X-ray baggage scanners. However, manual detection of contraband items is inherently challenging and time-consuming resulting in significant delays at crowded places such as airports, train-stations, shopping malls etc. Moreover, there is a significant risk of overlooking certain items that could pose potential harm. To address these challenges, there is a growing demand for intelligent systems for contraband items detection that can efficiently and accurately detect items whilst minimizing false negatives. Automated deep learning solutions offer a sophisticated and technologically advanced approach to enhance the accuracy and speed of the detection process. In our pursuit to address this challenge comprehensively, we have obtained an X-ray Imaging Dataset specifically curated for this purpose. The dataset includes five types of objects including guns, knives, pliers, scissors, and wrenches that are typically banned to carry along. In this paper, we have proposed a deep learning-based approach to efficiently and accurately detect contraband items from X-ray images. The proposed approach is based on YOLO architectures that has been shown to perform better for object detection in variety of domains both in terms of accuracy and real-time performance. We have evaluated different versions of YOLO to select the version that works best for contraband item detection from X-ray images. Yolo-v8 has shown superior performance followed by Yolo-v5 in terms of accuracy. Challenges regarding class imbalance have been addressed using data augmentation especially for clas

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Gradient-Free Methods for Nonconvex Nonsmooth Stochastic Compositional Optimization 38

Gradient-Free Methods for Nonconvex Nonsmooth Stochastic Com...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Liu, Zhuanghua Luo, Luo Low, Bryan Kian Hsiang Department of Computer Science National University of Singapore CNRS@CREATE Ltd. 1 Create Way #08-01 CREATE Tower 138602 Singapore School of Data Science Fudan University Shanghai Key Laboratory for Contemporary Applied Mathematics China Department of Computer Science National University of Singapore Singapore

Stochastic compositional optimization (SCO) problems are popular in many real-world applications, including risk management, reinforcement learning, and meta-learning. However, most of the previous methods for SCO require the smoothness assumption on both the outer and inner functions, which limits their applications to a wider range of problems. In this paper, we study the SCO problem in that both the outer and inner functions are Lipschitz continuous but possibly nonconvex and nonsmooth. In particular, we propose gradient-free stochastic methods for finding the (δ, ϵ)-Goldstein stationary points of such problems with non-asymptotic convergence rates. Our results also lead to an improved convergence rate for the convex nonsmooth SCO problem. Furthermore, we conduct numerical experiments to demonstrate the effectiveness of the proposed methods. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Testing Constraint Checking Implementations via Principled Metamorphic Transformations 31

Testing Constraint Checking Implementations via Principled M...

引用

31st IEEE International Conference on Software Analysis, Evolution and Reengineering, SANER 2024

作者： Gao, Mingchen Wang, Huiyan Xu, Chang Nanjing University State Key Laboratory for Novel Software Technology Nanjing China Nanjing University Department of Computer Science and Technology Nanjing China

ISBN: (纸本)9798350330663

Constraint checking techniques are being widely used for ensuring the consistency of software artifacts during their development and evolution (e.g., detecting inconsistency in an application's running contexts or identifying rule violation in the code being developed). Typically, consistency constraints are formulated and checked upon the changes of software artifacts under checking. When any constraint is violated, an inconsistency is said to occur and then follow-up actions can be taken to remedy the problem. Currently, various constraint checking techniques have been proposed and implemented with sophisticated mechanisms for higher efficiency and scalability. However, these implementations could be far from being well tested due to their oracle problems, i.e., hardly able to tell what the checking result should be, given any software artifacts and their consistency constraints to check. In this paper, we leverage metamorphic testing and propose a family of metamorphic relations catered for testing constraint checking implementations. We dedicatedly design these relations by following the sensitivity principle via a fine-granularity control and the diversity principle via input-oriented transformations. Our experiments reported promising results (disclosing 80 % mutation bugs and five real bugs) without the need of any manual labeling. © 2024 IEEE.

关键词： Software testing

来源：评论

学校读者我要写书评

暂无评论

Mjölnir: Breaking the Shield of Perturbation-Protected Gradients via Adaptive Diffusion 39

Mjölnir: Breaking the Shield of Perturbation-Protected Grad...

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Liu, Xuan Cai, Siqi Zhou, Qihua Guo, Song Li, Ruibin Lin, Kaiwei The Hong Kong Polytechnic University Hong Kong Hubei Key Laboratory of Transportation Internet of Things Wuhan University of Technology Wuhan China College of Computer Science and Software Engineering Shenzhen University Shenzhen China Department of Computer Science and Engineering The Hong Kong University of Science and Technology Hong Kong

ISBN: (纸本)157735897X

Perturbation-based mechanisms, such as differential privacy, mitigate gradient leakage attacks by introducing noise into the gradients, thereby preventing attackers from reconstructing clients’ private data from the leaked gradients. However, can gradient perturbation protection mechanisms truly defend against all gradient leakage attacks? In this paper, we present the first attempt to break the shield of gradient perturbation protection in Federated Learning for the extraction of private information. We focus on common noise distributions, specifically Gaussian and Laplace, and apply our approach to DNN and CNN models. We introduce Mjölnir, a perturbation-resilient gradient leakage attack that is capable of removing perturbations from gradients without requiring additional access to the original model structure or external data. Specifically, we leverage the inherent diffusion properties of gradient perturbation protection to develop a novel diffusion-based gradient denoising model for Mjölnir. By constructing a surrogate client model that captures the structure of perturbed gradients, we obtain crucial gradient data for training the diffusion model. We further utilize the insight that monitoring disturbance levels during the reverse diffusion process can enhance gradient denoising capabilities, allowing Mjölnir to generate gradients that closely approximate the original, unperturbed versions through adaptive sampling steps. Extensive experiments demonstrate that Mjölnir effectively recovers the protected gradients and exposes the Federated Learning process to the threat of gradient leakage, achieving superior performance in gradient denoising and private data recovery. Copyright © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Historical Analysis of Financial Fraud and Its Future 1

引用

International Conference on Applied Soft Computing and Communication Networks, ACN 2023

作者： Girish, K.K. Bhowmik, Biswajit Department of Computer Science and Engineering Ishwarchandra Vidyasagar AIT Lab BRICS Laboratory National Institute of Technology Karnataka Surathkal Mangalore575025 India

ISBN: (数字)9789819720040

ISBN: (纸本)9789819720033

As the world is sailing toward a highly advanced digital financial culture with the advent of financial technologies (FinTech), more and more people are now under the shore of financial inclusion. Subsequently, new opportunities are created in the financial sector, making it possible for people who were previously unbanked or underbanked to access financial services. However, the rise in financial fraud and its potential implications are creating a rift in the financial sector, resulting in substantial economic losses across the globe. This paper provides an in-depth comprehension of financial fraud, encompassing its historical perspectives and ramifications. After that, factors contributing to fraudulent behavior are highlighted. In addition, the paper presents a comprehensive framework for fraud classification and accentuates the impacts of financial fraud. Furthermore, the paper underscores the aspects influencing future occurrences of financial fraud, enabling the formulation of proactive strategies to prevent and mitigate it. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.

关键词： Losses

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：