检索结果-内蒙古大学图书馆

arXiv 2023年

作者： Pal, Ambar Sulam, Jeremias Department of Computer Science Mathematical Institute of Data Science Johns Hopkins University BaltimoreMD21218 United States

Randomized smoothing is a technique for providing provable robustness guarantees against adversarial attacks while making minimal assumptions about a classifier. This method relies on taking a majority vote of any base classifier over multiple noise-perturbed inputs to obtain a smoothed classifier, and it remains the tool of choice to certify deep and complex neural network models. Nonetheless, non-trivial performance of such smoothed classifier crucially depends on the base model being trained on noise-augmented data, i.e., on a smoothed input distribution. While widely adopted in practice, it is still unclear how this noisy training of the base classifier precisely affects the risk of the robust smoothed classifier, leading to heuristics and tricks that are poorly understood. In this work we analyze these trade-offs theoretically in a binary classification setting, proving that these common observations are not universal. We show that, without making stronger distributional assumptions, no benefit can be expected from predictors trained with noise-augmentation, and we further characterize distributions where such benefit is obtained. Our analysis has direct implications to the practical deployment of randomized smoothing, and we illustrate some of these via experiments on CIFAR-10 and MNIST, as well as on synthetic datasets. © 2023, CC BY.

关键词： Economic and social effects

来源：评论

学校读者我要写书评

暂无评论

Negative Adversarial Example Generation Against Naver's Celebrity Recognition API 1

Negative Adversarial Example Generation Against Naver's Cele...

引用

1st ACM Workshop on Security Implications of Deepfakes and Cheapfakes, WDC 2022, co-located with ACM AsiaCCS 2022

作者： Kim, Keeyoung Woo, Simon S. State University of New York Korea Republic of Stony Brook University & IngenioAI Seoul Korea Republic of Department of Artificial Intelligence Applied Data Science Department Computer Science & Engineering Department Sungkyunkwan University Suwon Korea Republic of

ISBN: (纸本)9781450391788

Deep Neural Networks (DNNs) are very effective in image classification, detection and recognition due to a large number of available data. However, they can be easily fooled by adversarial examples and produce incorrect results, which can cause problems for many applications. In this work, we focus on generating adversarial images and exploring and assessing possible negative impacts caused by these examples. As a case study, we create adversarial images against Naver's celebrity recognition (NCR) API, as Naver is the leading machine learning APIs service provider in South Korea. We demonstrate that it is extremely easy to fool the online DNN-based APIs using adversarial examples and discuss possible negative impacts resulting from these adversarial examples. © 2022 ACM.

关键词： Image classification

来源：评论

学校读者我要写书评

暂无评论

Development of Support Vector Machine Method to Forecast Dengue-Endemic Areas

Development of Support Vector Machine Method to Forecast Den...

引用

2023 International Conference on Informatics, Multimedia, Cyber and Information Systems, ICIMCIS 2023

作者： Meileni, Hetty Ermatita Abdiansah Husni, Nyayu Latifah Desmawati Doctoral Program in Engineering Science Palembang30139 Indonesia Universitas Sriwijaya Faculty of Computer Science Palembang30139 Indonesia Palembang30139 Indonesia Politeknik Negeri Sriwijaya Palembang Department of Electrical Engineering Palembang30139 Indonesia Universitas Pembangunan Nasional Veteran Jakarta Faculty of Health Science Jakarta12450 Indonesia

ISBN: (纸本)9798350307306

The cases of dengue hemorrhagic fever (DHF) in Indonesia have increased significantly since 2020. data shown by the Central Statistics Agency, for example, in South Sumatra Province, there were 6,348 cases of DHF during that period. To suppress the case number of DHF, forecasting areas that are susceptible to this disease is urgently needed. This research aims to develop a machine learning method Support Vector Machine (SVM). The data used include area data, population density data, weather data, and dengue fever case data. The development of the SVM model is carried out using a machine learning method that combines regional data, population density data, weather data, and dengue case data. In the end, using the SVM method can be used to forecast dengue-endemic areas, so that the government and society can take further action to reduce the negative impact of DHF. © 2023 IEEE.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

data and Information Fusion Oriented Outline Driven Framework for Clause Extraction from Electronic Insurance Contracts 23

Data and Information Fusion Oriented Outline Driven Framewor...

引用

23rd IEEE International Conference on High Performance Computing and Communications, 7th IEEE International Conference on data science and Systems, 19th IEEE International Conference on Smart City and 7th IEEE International Conference on Dependability in Sensor, Cloud and Big data Systems and Applications, HPCC-DSS-SmartCity-DependSys 2021

作者： Duan, Zhifeng Xu, Bin Wu, Zhengwu Duan, Yucong Ren, Chengxiang China Judicial Big Data Institute Department of Politics and Law Beijing China Hainan University School of Computer Science and Technology Haikou China

ISBN: (纸本)9781665494571

In the justice trail of insurance contract disputes, a judge is required to quickly understand the main points involved in an insurance contract to find related pieces of evidence. The content of an insurance contract is usually standardized and organized to some degree. Researchers and engineers can use techniques including deep learning and rule engines to analyze insurance contracts. Typical tasks include contract outline identification and contract clause extraction. However, judicial applications are facing the challenge of robustness, interpretability, process transparency and continuous improvement as well as accuracy and recall rate. This paper proposed an outline driven framework to extract hierarchical clauses from multimodal electronic contracts by com-bining rules summarized by experts and deep learning methods to continuously discover and improve the rule base. In our approach, conflicts of outlines can be solved globally by combining basic outline rules, outline sequence sorting, an entropy based outline sequence filtering and a global fault tolerance strat-egy. A computer vision method is used to extract complete contract clauses by their spatial position relations. Results from judicial datasets show that the proposed method can quickly and effectively achieve high performance in recall and accuracy compared with purely deep learning algorithms and can be built up in a short period, which shows the practical value in rapid and long-term iterative development. © 2021 IEEE.

关键词： Deep learning Filtering Fault tolerant systems Insurance Robustness Iterative algorithms data mining

来源：评论

学校读者我要写书评

暂无评论

Digital Nudge Evaluation on COVID-19 tracing Application 9

Digital Nudge Evaluation on COVID-19 tracing Application

引用

9th International Conference on Information and Communication Technology, ICoICT 2021

作者： Sukmaningsih, Dyah Wahyu Abdurachman, Edi Purwandari, Betty Trisetyarso, Agung Bina Nusantara University BINUS Graduate Program-Computer Science Computer Science Department Jakarta Indonesia Faculty of Computer Science Universitas Indonesia Depok Indonesia

ISBN: (纸本)9781665404471

Nudge is considered as an intervention to change user behavior and influence decision-making. Mobile apps have become a part of our everyday life. In this pandemic era, governments use mobile apps' technology to control the spread of COVID-19 infection. Many governments implemented COVID-19 tracing apps, and citizens are forced to use the apps to record their contact tracing and alert them if they have contact with an infected person. The key to control the increasing spread of the virus also depends on citizens adhere to health protocol. Government has experience difficulty in enforcing the protocol without strict surveillance. Therefore, building awareness of risk is essential-this study attempt to design nudge intervention into COVID-19 tracing apps. From the evaluation of the original apps, four nudge interventions can be applied in the application. © 2021 IEEE.

关键词： COVID-19

来源：评论

学校读者我要写书评

暂无评论

Novel and Robust Energy Efficient Protocol for Authentication Purpose in IIoT

Novel and Robust Energy Efficient Protocol for Authenticatio...

引用

2023 IEEE International Conference on Research Methodologies in Knowledge Management, Artificial Intelligence and Telecommunication Engineering, RMKMATE 2023

作者： Kirubakaran, D. Velmurugan, S. Rekha, R. Sasirekha Saravanan, S. Panurangan, Raji Department of Computer Science and Engineering R.M.K. Engineering College Tamil Nadu Chennai India Department of Electronics and Communication Engineering T J S Engineering College Tamil Nadu Chennai India Department of Computer Science and Engineering Panimalar Engineering College Tamil Nadu Chennai India Department of Computer Science and Engineering Prince Shri Venkateshwara Padmavathy Engineering College Tamil Nadu Chennai600127 India Department of Artificial Intelligence and Data Science Engineering Jeppiaar Institute of Technology Tamil Nadu Chennai India Saveetha Engineering College Department of Electronics and Communication Engineering Saveetha Nagar Thandalam Tamil Nadu Chennai India

ISBN: (纸本)9798350305708

An developing technology called the Internet of Things (IoT) is anticipated to offer answers for several industrial industries. Wireless sensor networks (WSNs), a foundational IoT technology, may be utilized to gather the necessary environmental characteristics for certain applications. Security in WSN has grown to be a huge concern as a result of the sensor node's resource limitations and the wireless channel's open nature. To ensure the legitimacy of data access in WSN, authentication may be utilized as a fundamental security function. For various security needs, Chang and Le have presented two authentication techniques for WSN. However, their protocol has significant security and functionality flaws and is unable to provide appropriate mutual authentication. To address the shortcomings of earlier methods, we introduce a three-factor user authentication scheme for WSN. The proposed protocol's security is examined, and its performance, functionality, and security are contrasted with those of other relevant protocols. The comparison and simulation findings by NS-3 demonstrate how reliable and resource-efficient the suggested protocol is for IoT applications. © 2023 IEEE.

关键词： energy efficient Internet of Things Packet Delivery Ratio Public Key Infrastructure Wireless Sensor Networks

来源：评论

学校读者我要写书评

暂无评论

Implementing Catboost Algorithm for Allergen Cross Contamination Detection in Food Industry 1

Implementing Catboost Algorithm for Allergen Cross Contamina...

引用

1st IEEE International Conference on Emerging Research in Computational science, ICERCS 2023

作者： Sudharson, D. Kumar, D. Satheesh Dubey, Aman Kumar Kumar, B. Arun Vaishali, V. Balavedhaa, S. Kumaraguru College of Technology Artificial Intelligence and Data Science Tamil Nadu Coimbatore India Hindusthan College of Engineering &technology Computer Science and Engineering Tamil Nadu Coimbatore India Kumaraguru College of Technology Master of Bussiness Administration Tamil Nadu Coimbatore India Karpagam Academy of Higher Education Department of Cse Tamil Nadu Coimbatore India Kumaraguru College of Engineering Artificial Intelligence and Data Science Tamil Nadu Coimbatore India

ISBN: (纸本)9798350359756

In the food industry, allergen cross-contamination is a serious problem that necessitates innovative approaches to detection and prevention. In order to tackle this problem, this paper investigates the possibilities of machine learning (ML) techniques, focusing on CatBoost, XGBoost, and LightGBM in particular. Through an analysis of datasets from food processing establishments, the research develops prediction models that leverage on the distinct advantages of these machine learning methods. The models' comparative evaluations demonstrate how well they identify allergen cross-contamination risks, improving food safety protocols. This paper demonstrates the significant role of CatBoost, XGBoost, and LightGBM in strategically controlling allergen cross-contamination and offers insightful information about the applicability of ML. In the final analysis, these results provide a framework for the advancement of risk-mitigating intelligent systems that ensure consumer safety and support the industry's broader food safety regulations. © 2023 IEEE.

关键词： Food safety

来源：评论

学校读者我要写书评

暂无评论

BiGeoDepth: Leveraging Bi-Geometric Priors for Unsupervised Monocular Depth Estimation in Indoor Environments

引用

IEEE Transactions on Consumer Electronics 2025年

作者： Dong, Li Ren, Qingji Shi, Jianyang Zhang, Bowen Shenzhen Technology University College of Big Data and Internet Shenzhen518055 China Harbin Institute of Technology Department of Computer Science Shenzhen518055 China

Monocular depth estimation for indoor scenes remains challenging, particularly in textureless regions such as walls, floors, and ceilings, where photometric reconstruction losses provide weak supervision. Existing unsupervised approaches often fail to fully exploit the geometric regularities of indoor environments. To address these limitations, we propose BiGeoDepth, an unsupervised depth estimation framework that leverages comprehensive geometric constraints. By fully integrating planar and linear structures, we introduce a novel bi-consistent loss that ensures the correctness of planes and lines in 3D space and enforces their alignment with the principal orthogonal axes under the Manhattan-world assumption. This approach provides stronger supervision signals beyond photometric loss, leading to more accurate depth predictions. Extensive experiments on the NYU Depth V2 dataset demonstrate that BiGeoDepth significantly outperforms existing methods, reducing absolute relative error, root mean squared error, and logarithmic root mean squared error by 0.007, 0.017, and 0.003, respectively, compared to PLNet. Additionally, validation on the ScanNet dataset shows the generalization capability of BiGeoDepth, making it applicable to a wide variety of indoor scenes. © 1975-2011 IEEE.

关键词： Mean square error

来源：评论

学校读者我要写书评

暂无评论

A Crystal-Specific Pre-Training Framework for Crystal Material Property Prediction

arXiv

引用

arXiv 2023年

作者： Yu, Haomin Song, Yanru Hu, Jilin Guo, Chenjuan Yang, Bin Department of Computer Science Aalborg University Denmark School of Data Science & Engineering East China Normal University China

Crystal property prediction is a crucial aspect of developing novel materials. However, there are two technical challenges to be addressed for speeding up the investigation of crystals. First, labeling crystal properties is intrinsically difficult due to the high cost and time involved in physical simulations or lab experiments. Second, crystals adhere to a specific quantum chemical principle known as periodic invariance, which is often not captured by existing machine learning methods. To overcome these challenges, we propose the crystal-specific pre-training framework for learning crystal representations with self-supervision. The framework designs a mutex mask strategy for enhancing representation learning so as to alleviate the limited labels available for crystal property prediction. Moreover, we take into account the specific periodic invariance in crystal structures by developing a periodic invariance multi-graph module and periodic attribute learning within our framework. This framework has been tested on eight different tasks. The experimental results on these tasks show that the framework achieves promising prediction performance and is able to outperform recent strong baselines. Copyright © 2023, The Authors. All rights reserved.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

An Extended CRITIC Algorithm Based on Spherical Fuzzy Sets to Determine the Objective Weights of Decision-Makers

An Extended CRITIC Algorithm Based on Spherical Fuzzy Sets t...

引用

International Conference on Informatics and Computing (ICIC)

作者： Irvanizam Irvanizam Mahyuddin K. M. Nasution Tulus Tulus Erna Budhiarti Nababan Computer Science Doctoral Program Universitas Sumatera Utara Medan Indonesia Department of Computer Science Universitas Sumatera Utara Medan Indonesia Department of Mathematics Universitas Sumatera Utara Medan Indonesia

ISBN: (数字)9798331517601

ISBN: (纸本)9798331517618

Determining decision-maker weight coefficients is a crucial step in multi-criteria group decision-making models. However, much of the existing literature focuses on developing algorithms to select the best alternative based on exact criteria weights, often leading to relatively subjective outcomes. Therefore, this study proposes an algorithm that objectively calculates decision-maker weight coefficients by extending the CRiteria Importance Through Interpreters Correlation (CRITIC) method within Spherical Fuzzy Sets (SFSs). The algorithm modifies the normalization process of the spherical fuzzy number (SFN) elements in the initial decision-maker priority matrix and enhances the SFN aggregation in the normalized priority matrix. By applying a distance measure between SFN elements, the algorithm reduces discrepancies between normalized elements, providing a more objective representation of the relationships in the initial priority matrix and leading to more accurate decision-maker weights. We validate the proposed algorithm through three case studies, demonstrating its feasibility and effectiveness in calculating objective decision-maker weights. In addition, the results indicate that the algorithm can be integrated with other decision-making methods, paving the way for the future development of dynamic decision-support systems.

关键词： Weight measurement Fuzzy sets Correlation Accuracy Heuristic algorithms Decision making Informatics Standards

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：