检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Khouna, Awa Ferry, Julien Vidal, Thibaut CIRRELT SCALE-AI Chair in Data-Driven Supply Chains Department of Mathematics and Industrial Engineering Poly-technique Montréal Canada

The advent of Machine Learning as a Service (MLaaS) has heightened the trade-off between model explainability and security. In particular, explainability techniques, such as counterfactual explanations, inadvertently increase the risk of model extraction attacks, enabling unauthorized replication of proprietary models. In this paper, we formalize and characterize the risks and inherent complexity of model reconstruction, focusing on the "oracle" queries required for faithfully inferring the underlying prediction function. We present the first formal analysis of model extraction attacks through the lens of competitive analysis, establishing a foundational framework to evaluate their efficiency. Focusing on models based on additive decision trees (e.g., decision trees, gradient boosting, and random forests), we introduce novel reconstruction algorithms that achieve provably perfect fidelity while demonstrating strong anytime performance. Our framework provides theoretical bounds on the query complexity for extracting tree-based model, offering new insights into the security vulnerabilities of their deployment. Copyright © 2025, The Authors. All rights reserved.

关键词： Decision trees

来源：评论

学校读者我要写书评

暂无评论

An explainable artificial intelligence – human collaborative model for investigating patent novelty

引用

engineering Applications of Artificial Intelligence 2025年 154卷

作者： Jang, Hyejin Yoon, Byungun Division of AI Data Convergence Hankuk University of Foreign Studies Gyeonggi-do Korea Republic of Department of Industrial & Systems Engineering Dongguk University Seoul Korea Republic of

With the accumulation of technology-related big data, including the patent database, existing studies have proposed a framework for patent analysis using natural language processing models. Artificial intelligence (ai) applications require human experience and insight based on the understanding of complex environments and uncertainties and model predictive performance. However, existing research has focused on applying big data and developing automated processes. Actual user understanding and the consideration of model usability are insufficient. Studies must consider the human–machine cooperation-based approach in developing the ai model. This study proposes a collaborative approach through which the explainable ai (Xai) model, a self-explaining deep neural network for text classification, communicates with users. The proposed Xai model provides users with an explanation for the model prediction along with the prediction results for patent evaluation. Users provide feedback based on the model predictions and their explanations. The source Xai model is refined via relearning by reflecting on user feedback. This study experiments to assess model improvement using the human collaboration method. As for the human collaborative method, this study considers the process of human intervention independent of the Xai model's results as well as the method of human participation based on the explanation presented by the Xai model. The experimental results verified the Xai model performance, showing the highest accuracy (0.890) and F1 score (0.916), such that the model can be applied efficiently to patent evaluation. The Xai–human collaboration model presented in this study can also be expanded and applied to technology intelligence tasks. However, the collaborative approach in this study has complete trust in human advice from technical experts;thus, subsequent collaborative Xai models could be improved by communicating bidirectionally with human resources as a complementary r

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Domain Separation aided data Efficiency Enhancement for Thermal Object Detection

Domain Separation Aided Data Efficiency Enhancement for Ther...

引用

International Conference on Electronics, Information and Communications (ICEIC)

作者： Beomsik Cho Seoyeong Choi Soomok Lee Department of Data Networks and AI Ajou University Suwon South Korea Department of Mechanical Engineering Ajou University Suwon South Korea Department of Mobility Engineering Ajou University Suwon South Korea

ISBN: (数字)9798331510756

ISBN: (纸本)9798331510763

Object detection is a critical component in autonomous driving systems, requiring robust performance across diverse lighting conditions, including nighttime scenarios where RGB cameras underperform due to low visibility. Thermal imaging overcomes this limitation by capturing infrared radiation, facilitating reliable object detection in low-light conditions. However, the development of deep learning models for thermal object detection is hindered by the scarcity of large-scale annotated thermal datasets. To tackle this challenge, we propose a novel approach that enhances data efficiency for thermal object detection through domain separation techniques. Our method introduces a Domain Invariant Encoder (DIE) and a Domain Specific Encoder (DSE). The DIE maps thermal images into a shared feature space with RGB images by minimizing the $\mathcal{H}$ -divergence between their feature distributions via adversarial learning. This alignment facilitates the transfer of knowledge from pre-trained RGB models to the thermal domain. Concurrently, the DSE captures unique thermal characteristics by learning domain-specific features. We integrate the invariant and specific features using a cross-attention mechanism, allowing the model to leverage shared and unique information effectively. Evaluated on the FLIR ADAS dataset, our approach achieves state-of-the-art performance with limited thermal data, demonstrating significant improvements in mean Average Precision (mAP) over baseline models. This study addresses the challenge of limited annotated data for thermal object detection, advancing thermal imaging applications in autonomous driving.

关键词： Measurement Deep learning Lighting Object detection Feature extraction data models Safety Reliability data mining Autonomous vehicles

来源：评论

学校读者我要写书评

暂无评论

Parallelizing Multi-objective A* Search

arXiv

引用

arXiv 2025年

作者： Ahmadi, Saman Sturtevant, Nathan R. Raith, Andrea Harabor, Daniel Jalili, Mahdi School of Engineering RMIT University Australia Department of Computing Science University of Alberta Canada Department of Engineering Science University of Auckland New Zealand Department of Data Science and AI Monash University Australia

The Multi-objective Shortest Path (MOSP) problem is a classic network optimization problem that aims to find all Pareto-optimal paths between two points in a graph with multiple edge costs. Recent studies on multi-objective search with A* (MOA*) have demonstrated superior performance in solving difficult MOSP instances. This paper presents a novel search framework that allows efficient parallelization of MOA* with different objective orders. The framework incorporates a unique upper bounding strategy that helps the search reduce the problem’s dimensionality to one in certain cases. Experimental results demonstrate that the proposed framework can enhance the performance of recent A*-based solutions, with the speed-up proportional to the problem dimension. Copyright © 2025, The Authors. All rights reserved.

关键词： Pareto principle

来源：评论

学校读者我要写书评

暂无评论

Recent Advances in Artificial Intelligence-Driven Prognostics and Health Management of Mobility Batteries

Multiscale Science and Engineering

引用

Multiscale Science and engineering 2025年 1-16页

作者： Kumar, Prashant Kim, Heung Soo Kim, Sungjun Kim, Hyun-Seok Moon, Janghyuk Department of AI and Big Data Woosong University Daejeon Republic of Korea Department of Mechanical Robotics and Energy Engineering Dongguk University Seoul Republic of Korea Division of Electronics and Electrical Engineering Dongguk University Seoul Republic of Korea Department of Energy Systems Engineering Chung-Ang University Seoul Republic of Korea

The rise in demand for sustainable transportation solutions in recent years has resulted in Electric vehicles (EVs) attracting substantial interest. The rapid rise in EV adoption has led to the need for efficient prognostics and health management (PHM) solutions. To ensure reliable and safe operation, the batteries that power EVs require maintenance. This paper reviews the detailed artificial intelligence (ai)-driven PHM framework for mobility batteries. Different mobility batteries have different chemistries owing to the different battery materials used for their development. ai-driven PHM uses machine learning, deep learning, and data-driven techniques to accurately estimate key battery parameters that include state of health and remaining useful life. We review data collection strategies, data processing approaches, feature development, and battery health assessment techniques based on ai, reviewing various ai methodologies that include artificial neural networks, Gaussian process regression, convolutional neural networks, and many more. This inclusive review of the recent trends and methodologies of ai-based mobility battery PHM provides a framework to develop a future efficient PHM solution for safer, efficient, and reliable battery systems.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Multi-resolution Score-Based Variational Graphical Diffusion for Causal Disaster System Modeling and Inference

arXiv

引用

arXiv 2025年

作者： Li, Xuechun Gao, Shan Xu, Susu Center for Systems Science and Engineering Department of Civil and Systems Engineering Johns Hopkins University United States Data Science and AI Institute Johns Hopkins University United States

Complex systems with intricate causal dependencies present significant challenges for accurate estimation and prediction. Effective modeling of these systems requires precise modeling of the underlying physical processes, integration of multiple interdependent factors, and incorporation of multi-source observational data with varying resolutions. These systems can manifest in both static scenarios with instantaneous causal chains and temporal scenarios with evolving dynamics, further complicating the modeling challenge. Existing methods struggle to simultaneously handle varying resolutions, capture physical relationships, model causal dependencies, and incorporate temporal dynamics, especially when data comes from diverse sources with inconsistent sampling. Therefore, we introduce Temporal-SVGDM: Score-based Variational Graphical Diffusion Model for Multi-resolution observations, which addresses these challenges through a novel integration of score-based diffusion models and causal graphical models. Our framework first constructs individual SDEs for each variable at its corresponding native resolution, then couples these SDEs through a causal score mechanism where parent nodes inform the evolution of the child nodes. This approach enables unified modeling of both immediate causal effects in static scenarios and evolving dependencies in temporal scenarios. In temporal models, these state representations are then processed through a temporal sequence prediction model to predict future states based on historical patterns and causal relationships. Through extensive experiments on real-world datasets, we demonstrate not only improved prediction accuracy and causal understanding compared to existing methods, but also robust performance under varying levels of background knowledge availability. Our model exhibits graceful degradation patterns across different disaster types, successfully handling both static earthquake scenarios and temporal hurricane and wildfire scenario

关键词： Prediction models

来源：评论

学校读者我要写书评

暂无评论

Multi-class Seismic Building Damage Assessment from InSAR Imagery using Quadratic Variational Causal Bayesian Inference

arXiv

引用

arXiv 2025年

作者： Li, Xuechun Xu, Susu Center for Systems Science and Engineering Department of Civil and Systems Engineering Johns Hopkins University United States Data Science and AI Institute Johns Hopkins University United States

Interferometric Synthetic Aperture Radar (InSAR) technology leverages satellite radar signals to detect surface deformation patterns, allowing large-scale monitoring of earthquake impacts on the built environment. These measurements can reveal structural responses of buildings to seismic forces, making InSAR valuable for rapid damage assessment. Although detailed multi-class building damage classifications (e.g., no damage, moderate, severe, collapse) are essential for emergency response planning during the critical first 72 hours, extracting this information from InSAR data presents significant challenges: the overlapping of building damage signatures with secondary hazards and environmental noise, computational intractability due to the exponential growth in model parameters in multi-class scenarios, and the need for rapid processing at regional scales. We address these challenges through a novel multi-class variational causal Bayesian inference framework with quadratic variational bounds - an approach that provides mathematically rigorous approximations of complex damage patterns while ensuring computational efficiency. By integrating InSAR observations with USGS ground failure models and building fragility functions, our framework effectively separates building damage signals while maintaining computational efficiency through strategic pruning. Evaluation across five major earthquakes (Haiti 2021, Puerto Rico 2020, Zagreb 2020, Italy 2016, Ridgecrest 2019) demonstrates substantial improvements in damage classification accuracy (AUC: 0.94-0.96), achieving up to 35.7% improvement over existing approaches. Our method shows particular robustness in areas with multiple overlapping hazard signatures, maintaining high accuracy (AUC ≥ 0.93) across all damage categories while reducing computational overhead by over 40% through strategic pruning, all without requiring extensive ground truth data. © 2025, CC BY.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Advanced fault diagnosis in industrial robots through hierarchical hyper-laplacian priors and singular spectrum analysis

引用

Complex and Intelligent Systems 2025年第6期11卷 1-26页

作者： Algburi, Riyadh Nazar Ali Aljibori, Hakim S. Sultan Al-Huda, Zaid Gu, Yeong Hyeon Al-antari, Mugahed A. Department of Mechanical Power Techniques Engineering Refrigeration and Air- Conditioning Al-Farahidi University Baghdad Iraq College of Engineering Mechanical Engineering University of Warith Al-Anbiyaa Karbala Karbala Iraq Stirling College Chengdu University Sichuan Chengdu 610106 China Department of Artificial Intelligence and Data Science College of AI Convergence Daeyang AI Center Sejong University Seoul 05006 South Korea

In industrial cases, robustness of the robots is mandatory and thus the development of fault diagnosis systems is essential. This study introduces a novel fault diagnosis method that merges two elements: Two methods shared here are the hierarchical hyper-Laplacian prior (HHLP) and singular spectrum analysis (SSA). The SSA technique decomposes the encoder signals into three components;residual, periodic oscillation and trend. In addition, the HHLP algorithm can identify harmonic interference, periodical impulses, and noise, with maximal posterior probabilities compared to the other algorithms. Compared to traditional Laplacian prior models, this approach provides higher accuracy, which verify the HHLP algorithm can effectively extract fault feature. Real-world applications and some computational studies provide additional light on the practicability of SSA-HHLP method. The research also compares the results with kurtosis-based weighted sparse prototypes, spectral kurtosis, and minimax concave regularization, and indicates that the proposed SSA-HHLP method outperforms other methods in both low outlier and high outlier contamination. © The Author(s) 2025.

关键词： Fault diagnosis Industrial robot Multi-scale cyclic modulation intensity Singular spectrum analysis Time series

来源：评论

学校读者我要写书评

暂无评论

Spatio-Temporal Machine Learning Approaches for Bus Travel Time Prediction

引用

Procedia Computer Science 2025年 257卷 553-560页

作者： Gokull Subramanian Aditya Verma Lelitha Vanajakshi Wadhwani School of Data Science & AI Indian Institute of Technology Madras India School of Civil Engineering Vellore Institute of Technology India Professor Department of Civil Engineering Indian Institute of Technology Madras India Robert Bosch Centre for Data Science and Artificial Intelligence (RBCDSAI) Indian Institute of Technology Madras India

Predicting bus travel times is critical for enhancing passenger satisfaction, optimizing route planning, and improving overall system efficiency. This study explores spatio-temporal modeling techniques to predict travel times for a bus route in Chennai, India, leveraging spatial dependencies across route sections and temporal variations in travel conditions. Machine learning algorithms, including Weighted K-Nearest Neighbors (KNN) and clustering-based approaches, were compared with deep learning methods like Recurrent Neural Networks (RNN). The models incorporated handcrafted spatio-temporal features and real-time data to improve prediction accuracy. The results reveal that machine learning models with spatio-temporal feature engineering performed competitively with deep learning models utilizing real-time data. The study presents the strengths and limitations of each approach, offering insights into effective spatio-temporal modeling for bus travel time prediction.

关键词： travel time prediction KNN clustering RNN spatio-temporal modeling

来源：评论

学校读者我要写书评

暂无评论

Efficient Traffic Density Calculation for Diverse Road Traffic using Yolov8 Algorithm

Efficient Traffic Density Calculation for Diverse Road Traff...

引用

International Conference on Communication Systems and Network Technologies (CSNT)

作者： Kotla Amulya Mandarapu Sahithya Mulka Trisha Shanmugasundaram Hariharan A V Senthil Kumar R. Karthikeyan Department of Artificial Intelligence and Data Science Vardhaman College of Engineering Hyderabad India Department of MCA Hindusthan College of Arts & Science Coimbatore Tamilnadu India Department of CSE(AI&ML) Vardhaman College of Engineering Hyderabad India

ISBN: (数字)9798331531935

ISBN: (纸本)9798331531942

Accurate traffic density estimation is vital for effective traffic management. Traditional methods like inductive loop detectors and radar have high costs and scalability issues. This report introduces a method using the YOLO (You Only Look Once) object detection algorithm to calculate traffic density. YOLOv3, fine-tuned with a custom dataset, enhances vehicle detection accuracy. The approach involves collecting video feeds from traffic surveillance cameras, sampling frames at regular intervals, and processing them with YOLOv3 to detect and count the number of vehicles. Traffic density is calculated based on vehicles per unit length of the road. Testing across various lighting conditions and traffic densities showed a detection rate consistently above 95%, validating the method's robustness. YOLO's real-time processing ensures efficient traffic density computation, supporting timely traffic control *** algorithm is also detects the vehicles but not accurate as YOLO algorithm as it does not detect the fast moving vehicles which leads to the inefficient use of algorithm.

关键词： YOLO Computer vision Accuracy Vehicle detection Surveillance Urban planning Traffic control Cameras Real-time systems Testing

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：