检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Qiao, Dan Zhang, Kaiqi Singh, Esha Soudry, Daniel Wang, Yu-Xiang Department of Computer Science UC Santa Barbara United States Technion – Israel Institute of Technology Halıcıoğlu Data Science Institute UC San Diego United States

We study the generalization of two-layer ReLU neural networks in a univariate nonparametric regression problem with noisy labels. This is a problem where kernels (e.g. NTK) are provably sub-optimal and benign overfitting does not happen, thus disqualifying existing theory for interpolating (0-loss, global optimal) solutions. We present a new theory of generalization for local minima that gradient descent with a constant learning rate can stably converge to. We show that gradient descent with a fixed learning rate η can only find local minima that represent smooth functions with a certain weighted first order total variation bounded by 1/η−1/2+Oe(σ+ √MSE) where σ is the label noise level, MSE is short for mean squared error against the ground truth, and Oe(·) hides a logarithmic factor. Under mild assumptions, we also prove a nearly-optimal MSE bound of Oe(n−4/5) within the strict interior of the support of the n data points. Our theoretical results are validated by extensive simulation that demonstrates large learning rate training induces sparse linear spline fits. To the best of our knowledge, we are the first to obtain generalization bound via minima stability in the non-interpolation case and the first to show ReLU NNs without regularization can achieve near-optimal rates in nonparametric regression. Copyright © 2024, The Authors. All rights reserved.

关键词： Interpolation

来源：评论

学校读者我要写书评

暂无评论

Learning Unbiased Risk Prediction Based Algorithms in Healthcare: A Case Study with Primary Care Patients

SSRN

引用

SSRN 2024年

作者： Gupta, Vibhuti Broughton, Julian Rukundo, Ange Pinky, Lubna Department of Biomedical Data Science School of Applied Computational Sciences Meharry Medical College 1005 Dr DB Todd Jr Blvd Nashville 37208 Tennessee USA Department of Computer Science and Data Science School of Applied Computational Sciences Meharry Medical College 1005 Dr DB Todd Jr Blvd Nashville 37208 Tennessee USA Department of Computer Science and Data Science School of Applied Computational Sciences Meharry Medical College 1005 Dr DB Todd Jr Blvd Nashville 37208 Tennessee USA Department of Biomedical Data Science School of Applied Computational Sciences Meharry Medical College 1005 Dr DB Todd Jr Blvd Nashville 37208 Tennessee USA

The proliferation of Artificial Intelligence (AI) has revolutionized the healthcare domain with technological advancements in conventional diagnosis and treatment methods. These advancements lead to faster disease detection, management and provide personalized healthcare solutions. However, most of the clinical AI methods developed and deployed in hospitals have algorithmic and data-driven biases due to insufficient representation of specific race, gender, and age group which leads to misdiagnosis, disparities, and unfair outcomes. The objective of this paper is to identify the potential sources of bias and mitigate them in healthcare data. We have selected an open source dataset of 48,785 primary care patients enrolled in risk-based contracts from 2013 to 2015 based on variety of variables including chronic conditions, medications, comorbidity, and healthcare cost. Through experiments with various machine learning algorithms leveraging Fairlearn tool, we have identified biases in the dataset, and then compare the impact of these biases on the prediction performance. Our results indicate that gender, racial, age, and intersectional bias have profound impact on the prediction performance which leads to unfair outcomes. © 2024, The Authors. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Harmony in the Australian Domain Space

arXiv

引用

arXiv 2024年

作者： Gong, Xian McCarthy, Paul X. Rizoiu, Marian-Andrei Boldi, Paolo University of Technology Sydney Sydney Australia The Data Science Institute University of New South Wales Sydney Australia Computer Science Department Università degli Studi di Milano Milan Italy

In this paper we use for the first time a systematic approach in the study of harmonic centrality at a Web domain level, and gather a number of significant new findings about the Australian web. In particular, we explore the relationship between economic diversity at the firm level and the structure of the Web within the Australian domain space, using harmonic centrality as the main structural feature. The distribution of harmonic centrality values is analyzed over time, and we find that the distributions exhibit a consistent pattern across the different years. The observed distribution is well captured by a partition of the domain space into six clusters;the temporal movement of domain names across these six positions yields insights into the Australian Domain Space and exhibits correlations with other non-structural characteristics. From a more global perspective, we find a significant correlation between the median harmonic centrality of all domains in each OECD country and one measure of global trust, the WJP Rule of Law Index. Further investigation demonstrates that 35 countries in OECD share similar harmonic centrality distributions. The observed homogeneity in distribution presents a compelling avenue for exploration, potentially unveiling critical corporate, regional, or national insights. © 2024, CC BY.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Trace-Selection Preprocessing Approach of Discovering Structurally Complete ICN-Process Models from Process Logs 3rd

A Trace-Selection Preprocessing Approach of Discovering Str...

引用

3rd International Conference on Advances in Information and Communication Technology, ICTA 2024

作者： Pham, Dinh-Lam Batzaya, Tsagaantsooj Kim, Kyoung-Sook Chang, Jeong-Hyun Kim, Kwanghoon Pio CCSRI Kyonggi University Gyeonggi Suwon16227 Korea Republic of Department of Public Safety Bigdata Graduate School of Kyonggi University Gyeonggi Suwon16227 Korea Republic of Data Process Engineering Research Laboratory Faculty of AI Computer Science and Engineering Kyonggi University Gyeonggi Suwon16227 Korea Republic of

ISBN: (纸本)9783031809422

In business process life-cycle management and reengineering through process mining, it is crucial for the process mining system to discover structurally safe and complete business process models from process logs. However, most process mining systems typically suffer from discovering inaccurate and unstructured business process models due to various types of anomalous traces hidden in process logs. In this paper, we therefore propose a process log preprocessing approach that purifies a corresponding process log dataset by selecting a group of reasonable process log traces and applying to discover an essential business process model satisfying the structural completeness requirement of matched-pairing and proper-nesting properties. The theoretical basis of the proposed approach is the mathematical graph model of information control net model that is the typical representation of business process models. Finally, we practically demonstrate the conceptual excellence and functional correctness of the proposed preprocessing approach through experimental verifications using real business process enactment log datasets available in the 4TU Center for Research data. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Reengineering

来源：评论

学校读者我要写书评

暂无评论

Efficient Detection of Knee Osteoporosis Using the Swin Transformer on X-Ray Images

Efficient Detection of Knee Osteoporosis Using the Swin Tran...

引用

International Conference on computer and Applications (ICCA)

作者： Rahat Sultana Muhammad Zaman Nikola Ivković Korhan Cengiz Iqra Rafaqat Adnan Akhunzada Faculty of Computer Science The University of Lahore Sargodha Pakistan Department of Computer Science COMSATS University Islamabad Islamabad Pakistan Faculty of Organization and Informatics University of Zagreb Varaždin Croatia Department of E&E Engineering Biruni University Istanbul Turkey Department of Data and Cybersecurity University of Doha for Science and Technology Doha Qatar

ISBN: (数字)9798350367560

ISBN: (纸本)9798350367577

Knee Osteoporosis (KOP) is a skeletal disease that is caused by low bone mineral density and the degradation of bone tissue. It increases the risk of bone fractures in the knee region and is commonly seen in older people but usually ignored due to its silent nature. Osteoporosis is mostly diagnosed via X-ray images and Dual-Energy X-ray Absorptiometry (DXA) scans that can be difficult to read due to their sheer volume the subtle variations. This study aims to develop a Deep Learning (DL) utilizing Swin Transform model that can accurately recognize osteoporosis in knee X-ray images at an early stage. The dataset used for this research comprises knee X-rays labeled as either “normal” or “osteoporotic,” with image preprocessing including resizing to standard dimensions, converting the images into tensors, and normalizing pixel values to improve model performance. The model is trained and validated on arranged dataset, characterized by metrics such as accuracy, precision, recall or F1-score that gives an idea how well our model performs for prediction as acknowledged for its performance in image classification tasks. The results demonstrate that the Swin Transformer model has an overall accuracy of 89.38% on the data set for identifying osteoporosis features on knee X-ray images. This approach demonstrates the potential of the Swin Transformer model to serve as an effective tool for early osteoporosis diagnosis, offering improved accuracy and reliability compared to traditional convolutional neural networks (CNNs).

关键词： Deep learning Training Osteoporosis Accuracy Transforms Predictive models Transformers Feature extraction data models X-ray imaging

来源：评论

学校读者我要写书评

暂无评论

SERDA: Secure Enhanced and Robust data Aggregation Scheme for Smart Grid

引用

Journal of computers (Taiwan) 2024年第5期35卷 73-90页

作者： Chen, Yiming Duan, Hongjun Wang, Dong Liu, Yining Chang, Chin-Chen School of Computer Science and Information Security Guilin University of Electronic Technology Guilin541004 China Zhengzhou Art Infant Normal School Zhengzhou450000 China School of Mathematics and Computational Science Guilin University of Electronic Technology Guilin541004 China School of Data Science and Artificial Intelligence Wenzhou University of Technology Wenzhou325027 China Department of Information Engineering and Computer Science Feng Chia University Taichung40724 Taiwan Guangxi Colleges Universities Key Laboratory of Data Analysis and Computation Guilin541004 China

data aggregation is considered a viable security and privacy solution for smart grid as it allows to obtain the total electricity consumption within a region without disclosing individual data. However, existing data aggregation schemes give little consideration in their threat models to use cases where devices operate in untrustworthy environments and adversaries have physical system access, which is common in the smart grid. They cannot support authentication and resist physical attacks while maintaining data privacy and supporting fault tolerance for smart meter (SM) failures. Motivated by this, a secure enhanced and robust data aggregation (SERDA) scheme for smart grid is introduced in this article. The SERDA scheme provides enhanced security for key storage and updates based on physically unclonable function (PUF), while supporting data privacy protection and fault tolerance for SM failures without reliance on a trusted third party (TTP). Security analysis and performance evaluation demonstrate that SERDA meets the excepted goals and is efficient compared with related work. © 2024 Codon Publications. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Retinal Image Analysis for Early Deduction of Diabetic Retinopathy

Retinal Image Analysis for Early Deduction of Diabetic Retin...

引用

International Conference on Advanced Computing and Communication Systems (ICACCS)

作者： A. Chinnasamy P. Selvakumari Department of Data Science and Business Systems School of Computing SRMIST Chennai India Department of Computer Science and Engineering Chennai Institute of Technology Anna University Chennai India

The prevalence of DR is steadily rising, which necessitates the automatic disease severity extraction and classification. About 2% of those with this illness are entirely blind because of the diabetic mellitus problem and 10% get vision impairment after 15 years of diabetes as a result of the DR complication. It is also a significant contributor to blindness in both middle aged and older age groups. The patient may develop to severe stages of irreversible blindness if the condition is not detected early. The growing number of diabetic patients face a major issue due to a lack of ophthalmologists. It is suggested that an automated DR screening system be created to aid the ophthalmologist in making decisions. One of the primary symptoms of the DR is hard exudates. The detection of hard exudates is crucial for screening purposes and aids in disease monitoring and diagnosis. Thus, utilising Lloyd’s clustering technique, this work offered a unique techniques to segment the exudates and irregularities in DR were found.

关键词： Support vector machines Image segmentation Visual impairment Blindness Blood vessels Retina Time measurement

来源：评论

学校读者我要写书评

暂无评论

Productive Crop Field Detection: A New dataset and Deep Learning Benchmark Results

arXiv

引用

arXiv 2023年

作者： Nascimento, Eduardo Just, John Almeida, Jurandy Almeida, Tiago The Department of Computer Science Federal University of São Carlos São Paulo Brazil The Department of Data Science and Analytics John Deere IA United States

—In precision agriculture, detecting productive crop fields is an essential practice that allows the farmer to evaluate operating performance separately and compare different seed varieties, pesticides, and fertilizers. However, manually identifying productive fields is often time-consuming, costly, and subjective. Previous studies explore different methods to detect crop fields using advanced machine learning algorithms to support the specialists’ decisions, but they often lack good quality labeled data. In this context, we propose a high-quality dataset generated by machine operation combined with Sentinel-2 images tracked over time. As far as we know, it is the first one to overcome the lack of labeled samples by using this technique. In sequence, we apply a semi-supervised classification of unlabeled data and state-of-the-art supervised and self-supervised deep learning methods to detect productive crop fields automatically. Finally, the results demonstrate high accuracy in Positive Unlabeled learning, which perfectly fits the problem where we have high confidence in the positive samples. Best performances have been found in Triplet Loss Siamese given the existence of an accurate dataset and Contrastive Learning considering situations where we do not have a comprehensive labeled dataset available. Copyright © 2023, The Authors. All rights reserved.

关键词： Precision agriculture

来源：评论

学校读者我要写书评

暂无评论

Lightweight Pedestrian Detection through Guided Filtering and Deep Learning 10

Lightweight Pedestrian Detection through Guided Filtering an...

引用

10th IEEE Global Conference on Consumer Electronics, GCCE 2021

作者： Muchtar, Kahlil Saddami, Khairun Bintang, Akhyar Cenggoro, Tjeng Wawan Pardamean, Bens Lin, Chih-Yang Ernita, Tia Universitas Syiah Kuala Department of Electrical And Computer Engineering Aceh Indonesia Universitas Syiah Kuala Department of Electrical And Computer Engineering Aceh Indonesia Bina Nusantara University School of Computer Science Bina Nusantara University Bioinformatics And Data Science Research Center Computer Science Department Jakarta Indonesia Bina Nusantara University BINUS Graduate Program-Master of Computer Science Program Bina Nusantara University Bioinformatics And Data Science Research Center Computer Science Department Jakarta Indonesia Yuan-Ze University Department of Electrical And Engineering Taoyuan Taiwan

ISBN: (纸本)9781665436762

In this paper, we propose a novel approach to locate and detect moving pedestrians in a video. Our proposed method first locates the region of interest (ROI) using a background subtraction algorithm based on guided filtering. This novel background subtraction algorithm allows our method to also filter unexpected noises at the same time, which could benefit the performance of our proposed method. Subsequently, the pedestrians are detected using YOLOv3 within the provided ROI. Our experiments showed that the proposed method has a competitive performance in the CDNET2014 dataset with a fast-processing time. © 2021 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Hyperpersonalization of Educational Content Through Multimodal Deep Learning and Gamification

Hyperpersonalization of Educational Content Through Multimod...

引用

data science, Agents & Artificial Intelligence (ICDSAAI), International Conference on

作者： K. Saranya Veeramalai Sankaradass Department of Artificial Intelligence and Data Science Panimalar Institute of Technology Poonamalle Chennai India Department of Computer Science and Engineering Chennai Institute of Technology Kundrathur Tamilnadu India

This project proposes a groundbreaking approach to revolutionize the traditional educational framework by integrating multimodal deep learning and gamification to foster hyper-personalized learning experiences. Envisioning a classroom where education resonates with individual cognitive rhythms and passions, the system employs advanced algorithms to analyze text, voice, and facial expressions, thereby adapting content delivery to cater to diverse learning styles and needs. Concurrently, gamified elements such as points, badges, and personalized challenges enhance motivation, engagement, and retention. The teacher’s role transforms from a primary knowledge source to a skilled guide and facilitator, leveraging AI-driven insights to curate tailored learning experiences and interventions. This paradigm shift in pedagogy aims to create an inclusive, empowering, and adaptive learning environment where every student embarks on a personalized educational odyssey, celebrating their unique potential and fostering lifelong learning.

关键词：

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：