检索结果-内蒙古大学图书馆

An active learning framework for adversarial training of deep neural networks

Neural Computing and Applications 2025年第9期37卷 6849-6876页

作者： Ghosh, Susmita Chatterjee, Abhiroop Fiondella, Lance Department of Computer Science and Engineering Jadavpur University Kolkata India Department of Electrical and Computer Engineering University of Massachusetts Dartmouth United States

This article introduces a novel approach to bolster the robustness of Deep Neural Network (DNN) models against adversarial attacks named "Targeted Adversarial Resilience Learning (TARL)". The initial evaluation of a baseline DNN model reveals a significant accuracy decline when subjected to adversarial examples generated through techniques like FGSM, PGD, Carlini Wagner, and DeepFool attacks. To address this vulnerability, the article proposes an active learning framework, wherein the model iteratively identifies and learns from the most uncertain and misclassified instances. The key components of this approach include uncertainty estimation score in predicting the class of the input sample, selecting challenging samples based on this uncertainty score, labeling these challenging examples and augmenting them into the training set, and thereafter retraining the model with the expanded training set. The iterative active learning process, governed by parameters such as the number of iterations and batch size, demonstrates the potential to systematically enhance the resilience of DNN against adversarial threats. The proposed methodology has been investigated on several popular datasets such as the SARS-CoV-2 CT scan, MNIST, CIFAR-10, and Caltech-101, and demonstrated to be effective. Experiments illustrate that the learning framework improves the adversarial accuracies from 17.4% to 98.71% for the SARS-CoV-2 dataset, from 8.4% to 99.89% for the MNIST dataset, 1.6% to 78.84% for the CIFAR-10, and 12% to 92.92% for Caltech-101. Further, comparative analysis with several state-of-the-art methods suggests that the proposed framework offers superior defense against various attack methods and offers promising defensive mechanisms to deep neural networks. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Active learning

来源：评论

学校读者我要写书评

暂无评论

Deep Learning to Localize Photoacoustic Sources in Three Dimensions: Theory and Implementation

引用

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control 2025年第6期72卷 786-805页

作者： Gubbi, Mardava R. Lediju Bell, Muyinatu A. Johns Hopkins University Department of Electrical and Computer Engineering BaltimoreMD United States Johns Hopkins University Department of Electrical and Computer Engineering Department of Computer Science Department of Biomedical Engineering BaltimoreMD United States

Surgical tool tip localization and tracking are essential components of surgical and interventional procedures. The cross sections of tool tips can be considered as acoustic point sources to achieve these tasks with deep learning applied to photoacoustic channel data. However, source localization was previously limited to the lateral and axial dimensions of an ultrasound transducer. In this paper, we developed a novel deep learning-based three-dimensional (3D) photoacoustic point source localization system using an object detection-based approach extended from our previous work. In addition, we derived theoretical relationships among point source locations, sound speeds, and waveform shapes in raw photoacoustic channel data frames. We then used this theory to develop a novel deep learning instance segmentation-based 3D point source localization system. When tested with 4,000 simulated, 993 phantom, and 1,983 ex vivo channel data frames, the two systems achieved F1 scores as high as 99.82%, 93.05%, and 98.20%, respectively, and Euclidean localization errors (mean ± one standard deviation) as low as 1.46±1.11 mm, 1.58±1.30 mm, and 1.55±0.86 mm, respectively. In addition, the instance segmentation-based system simultaneously estimated sound speeds with absolute errors (mean ± one standard deviation) of 19.22±26.26 m/s in simulated data and standard deviations ranging 14.6-32.3 m/s in experimental data. These results demonstrate the potential of the proposed photoacoustic imaging-based methods to localize and track tool tips in three dimensions during surgical and interventional procedures. © 1986-2012 IEEE.

关键词： Ultrasonic transducers

来源：评论

学校读者我要写书评

暂无评论

DenseRSE-ASPPNet: An Enhanced DenseNet169 with Residual Dense Blocks and CE-HSOA-Based Optimization for IoT Botnet Detection

引用

International Journal of Advanced computer science and Applications 2025年第4期16卷 688-699页

作者： Khan, Mohd Abdul Rahim Department of Electrical Engineering and Computer Science A’sharqiyah University IBRA-400 Oman

The growing prevalence of Internet of Things (IoT) devices has heightened vulnerabilities to botnet-based cyberattacks, necessitating robust detection mechanisms. This paper proposes DenseRSE-ASPPNet, an advanced deep learning framework for botnet detection, incorporating comprehensive preprocessing, feature extraction, and optimization. The preprocessing pipeline includes data cleaning and Min-Max normalization to ensure high-quality input data. The DenseNet169 backbone is enhanced with Residual Squeeze-and-Excitation (RSE) blocks for channel-wise attention recalibration and Atrous Spatial Pyramid Pooling (ASPP) for capturing multi-scale spatial patterns, enabling effective feature extraction. Hyperparameter optimization is performed using the Cyclone-Enhanced Humboldt Squid Optimization Algorithm (CE-HSOA), which balances global exploration and local exploitation, ensuring faster convergence and enhanced robustness. Experimental results demonstrate the superior performance of the proposed framework, achieving 99.00 per cent accuracy, 96.40 per cent sensitivity, and 99.95 per cent specificity, significantly minimizing false positives and false negatives. The proposed DenseRSE-ASPPNet provides an efficient, scalable, and effective solution for mitigating botnet threats in IoT environments. © (2025), (science and Information Organization). All rights reserved.

关键词： Feature Selection

来源：评论

学校读者我要写书评

暂无评论

Maximizing aged photovoltaic array power:a computer modelling study

引用

Clean Energy 2024年第1期8卷 171-187页

作者： Fadi N.Sibai Electrical&Computer Engineering Department Gulf University for Science and TechnologyMishrefKuwait

Photovoltaic(PV)modules age with time for various reasons such as corroded joints and terminals and glass coating defects,and their ageing degrades the PV array *** the help of the PV array numerical model,this paper explores the effects of PV module ageing on the PV array power,and the power gains and costs of rearranging and recabling aged PV modules in a PV *** numerical PV array model is first revised to account for module ageing,rearrangement and recabling,with the relevant equations presented *** updated numerical model is then used to obtain the array powers for seven different PV *** power results are then analysed in view of the attributes of the seven PV array examples.A guiding method to recommend recabling after rearranging aged modules is then proposed,leading to further significant power gains,while eliminating intra-row *** certain conditions are met,it was shown that recabling PV modules after rearranging them may lead to further significant power gains,reaching 57%and 98%in two considered PV array *** gains are possible in other arrays.A cost-benefit analysis weighing annual power gains versus estimated recabling costs is also given for the seven considered PV array examples to guide recabling decisions based on technical and economic *** the considered examples,recabling costs can be recovered in<4 *** with the powers of the aged arrays,power gains due to our proposed rearranging and recabling the PV arrays ranged between 73%and 131%in the considered examples—well over the gains reported in the ***,the cost of our static module rearrangement and recabling method outshines the costs of dynamic reconfiguration methods recently published in the literature.

关键词： solar energy optimizing array power photovoltaic array modelling photovoltaic modules ageing photovoltaic module rearrangement module recabling and reconfiguration PV array mismatch

来源：评论

学校读者我要写书评

暂无评论

Statistical Process Monitoring Based on Ensemble Structure Analysis

引用

IEEE/CAA Journal of Automatica Sinica 2024年第8期11卷 1889-1891页

作者： Likang Shi Chudong Tong Ting Lan Xuhua Shi Faculty of Electrical Engineering and Computer Science Ningbo UniversityNingbo 315211China

Dear Editor,This letter presents a novel process monitoring model based on ensemble structure analysis(ESA).The ESA model takes advantage of principal component analysis(PCA),locality preserving projections(LPP),and multi-manifold projections(MMP)models,and then combines the multiple solutions within an ensemble result through Bayesian *** the developed ESA model,different structure features of the given dataset are taken into account simultaneously,the suitability and reliability of the ESA-based monitoring model are then illustrated through ***:The requirement for ensuring safe operation and improving process efficiency has led to increased research activity in the field of process monitoring.

关键词： ensemble preserving letter

来源：评论

学校读者我要写书评

暂无评论

A Semi-Tensor Product Based All Solutions Boolean Satisfiability Solver

引用

Journal of computer science & Technology 2023年第3期38卷 702-713页

作者：潘鸿洋储飞 Faculty of Electrical Engineering and Computer Science Ningbo UniversityNingbo 315211China

Boolean satisfiability (SAT) is widely used as a solver engine in electronic design automation (EDA). Typically, SAT is used to determine whether one or more groups of variables can be combined to form a true formula. All solutions SAT (AllSAT) is a variant of the SAT problem. In the fields of formal verification and pattern generation, AllSAT is particularly useful because it efficiently enumerates all possible solutions. In this paper, a semi-tensor product (STP) based AllSAT solver is proposed. The solver can solve instances described in both the conjunctive normal form (CNF) and circuit form. The implementation of our method differs from incremental enumeration because we do not add blocking conditions for existing solutions, but rather compute the matrices to obtain all the solutions in one pass. Additionally, the logical matrices support a variety of logic operations. Results from experiments with MCNC benchmarks using CNF-based and circuit-based forms show that our method can accelerate CPU time by 8.1x (238x maximum) and 19.9x (72x maximum), respectively.

关键词： all solutions Boolean satisfiability(AllSAT) semi-tensor product of matrices conjunctive normal form(CNF)satisfiability circuit satisfiability

来源：评论

学校读者我要写书评

暂无评论

QoS-Aware Application Assignment and Resource Utilisation Maximization Using AHP in Edge Computing

引用

IEEE Internet of Things Journal 2025年第11期12卷 17717-17728页

作者： Koganti, Yasasvitha Sridhar, Vidhyuth Yadav, Ram Narayan Pratap, Ajay Iitram Department of Electrical and Computer Science Engineering Ahmedabad380026 India Indiana University Department of Computer Science and Engineering Bloomington United States Banaras Hindu University Indian Institute of Technology Department of Computer Science and Engineering Varanasi India

Edge computing has emerged as a promising technology to satisfy the demand for data computational resources in Internet of Things (IoT) networks. With edge computing, processing of the massive data-intensive tasks can be done in the proximity of IoT users. Thus, the required constraints related to tasks such as latency requirements, and the Quality Of Service (QoS) can be guaranteed. However, the question that how to determine the task offloading strategy under various constraints of resources, distance, and cost is still an open issue. In this paper, we study the task offloading problem from a matching perspective and propose an Edge-User Assignment Algorithm (EUAA) that aims to maximize the resource utilization of edge servers and maximize the number of assigned IoT users. The main concern, in any matching algorithm, is how to generate the preference order of either side. To generate preference orders for edge servers, we use the concept of the Analytical Hierarchy Process (AHP). We have considered the following criteria: distance from users to the server, latency, available resources, and pricing. This generates the priority of the users for matching to edge servers. From IoT users' perspective, we use cost, and QoS parameters to improve their satisfaction. We have used state-of-the-art schemes to compare the performance of the proposed algorithm under different simulation scenarios. We compare the performance using the number of assigned users, servers' profit, number of satisfied users, edge server resources used, and execution time. The simulation results confirm that significant profit and satisfied user can be achieved by the proposed algorithm. © 2014 IEEE.

关键词： Computation offloading

来源：评论

学校读者我要写书评

暂无评论

PS-CoT-Adapter: adapting plan-and-solve chain-of-thought for scienceQA

引用

science China(Information sciences) 2025年第1期68卷 392-393页

作者： Qun LI Haixin SUN Fu XIAO Yiming WANG Xinping GAO Bir BHANU School of Computer Science Nanjing University of Posts and Telecommunications Purple Mountain Laboratories Department of Electrical and Computer Engineering University of California at Riverside

Large language models (LLMs) have recently shown remarkable performance in a variety of natural language processing (NLP) *** further explore LLMs'reasoning abilities in solving complex problems,recent research [1-3]has investigated chain-of-thought (CoT) reasoning in complex multimodal scenarios,such as science question answering (scienceQA) tasks [4],by fine-tuning multimodal models through human-annotated CoT ***,collected CoT rationales often miss the necessary rea-soning steps and specific expertise.

关键词：

来源：评论

学校读者我要写书评

暂无评论

FedQP: Large-Scale Private and Flexible Federated Query Processing

引用

IEEE Access 2025年 13卷 92019-92035页

作者： Almohri, Hussain M. J. Watson, Layne T. Kuwait University Department of Computer Science Kuwait City Kuwait Virginia Tech Departments of Computer Science Mathematics Aerospace and Ocean Engineering BlacksburgVA United States

State-of-the-art federated learning methods orchestrate iterations of the stochastic gradient descent algorithm among a network of clients to refine a unified set of model parameters, all while safeguarding individual client datasets. Although promising, the success of existing approaches hinges on using a uniform data model with the same optimization and learning algorithms by the participants. In addition, existing methods can be subject to privacy attacks, such as model-inversion attacks. A pivotal challenge lies in crafting an algorithm that can efficiently aggregate client inputs, ensuring minimal client dataset exposure and maximum flexibility in model adaptation. We propose a new scheme for aggregating local, diverse, and independent data models, addressing data inconsistency and model-inversion attacks. The proposed scheme, federated query processing FedQP, enables clients to build local models independently, without coordinating with the server or other clients. The server communicates with selected clients only to predict future values. Our scheme only requires one communication round between each client and the server for model initialization without the need for a training phase. Clients do not share data or model parameters. We evaluated FedQP by assessing the impact of our new aggregation mechanism on model accuracy and the communication efficacy of our architecture. We show that the caching algorithm in FedQP reduces redundant client communications during query processing. In addition, we present a privacy analysis that shows that our method outperforms prominent gradient-based approaches to federated learning. Our experiments show that FedQP consistently achieves higher classification accuracy in non-IID settings and demonstrates stronger resilience to reconstruction attacks compared to gradient-based methods such as federated averaging (FedAvg). © 2013 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Optimal Design of a Planning and Contracting Framework to Enable Vehicle to Building and Grid Services via Demand Response

引用

IEEE Transactions on Transportation Electrification 2025年第3期11卷 7542-7556页

作者： Elsayed, Ahmed Abd Elaziz Saxena, Shivam Farag, Hany E. Z. Electrical Engineering and Computer Science Department Lassonde School of Engineering York University Keele Campus Toronto ON Canada Department of Electrical and Computer Engineering University of New Brunswick Fredericton NB Canada

Bidirectional electric vehicle (EV) charging enables stored energy to reduce peak loads for buildings (V2B) and the grid (V2G). However, building owners investing in V2B infrastructure while generating revenue from V2G services face challenges in planning and coordinating with EV owners due to uncertainties in their schedules and profit-sharing expectations. Additionally, misaligned building and grid peak times can create conflicts between V2B and V2G goals, which may negatively impact the building electricity bill. Unlike previous studies that used a contract-free approach for aggregating V2B and V2G, resulting in inconsistent participation, this paper proposes a novel planning and contracting framework that enables building owner to determine the optimal contract parameters with EV owners. These parameters include minimum participation time in DR events, minimum arrival State of Charge (SoC), and permitted emergency departure hours. The framework supports V2B aggregation with on-site distributed energy resources for DR and V2G services, ensuring transparency and fairness through shared profits and performance-based penalties, while compensating building electricity bills due to V2G activities. The framework is composed of a tri-stage optimization process that uses Monte Carlo Simulations to generate EV owner profit assessments, select optimal EV candidates based on charger availability, estimate contract parameters with profit/penalty sharing, and assign contracts between power system operators, building owners, and EV owners under multiple virtual scenarios. Simulation results validate the contract parameter assessment and its performance via a three day case study with real-world data sets that demonstrates net revenue generation of $209 for each EV owner and $950 for building owners during DR events with a Return on Investment (ROI) of 134.5% 130.7% for the EVs and Building owners, respectively. © 2015 IEEE.

关键词： Vehicle-to-grid

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：