检索结果-内蒙古大学图书馆

IEEE Transactions on Artificial Intelligence 2023年第3期4卷 416-427页

作者： Su, DiJia Su, Difei Mulvey, John M. Poor, H.Vincent Princeton University Department of Electrical and Computer Engineering PrincetonNJ08540 United States University of British Columbia Department of Computer Science VancouverBCV6Z1T8 Canada Princeton University Department of Operation Research and Financial Engineering PrincetonNJ08540 United States

We consider extractive summarization within a cluster of related texts (multidocument summarization). Unlike single-document summarization, redundancy is particularly important because sentences across related documents might convey overlapping information. Thus, sentence extraction in such a setting is difficult because one will need to determine which pieces of information are relevant while avoiding unnecessary repetitiveness. To solve this difficult problem, we propose a novel reinforcement learning-based method Policy Blending with maximal marginal relevance and Reinforcement Learning (PoBRL) for solving multidocument summarization. PoBRL jointly optimizes over the following objectives necessary for a high-quality summary: importance, relevance, and length. Our strategy decouples this multiobjective optimization into different subproblems that can be solved individually by reinforcement learning. Utilizing PoBRL, we then blend each learned policies to produce a summary that is a concise and a complete representation of the original input. Our empirical analysis shows high performance on several multidocument datasets. Human evaluation also shows that our method produces high-quality output. © 2020 IEEE.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Message Passing Based Moving Target Localization in Multistatistic Systems With Unknown Time and Frequency Offsets

引用

IEEE Transactions on Aerospace and Electronic Systems 2025年

作者： Li, Wenjuan Jin, Ming Guo, Qinghua Ningbo University Faculty of Electrical Engineering and Computer Science Ningbo315211 China University of Wollongong School of Electrical Computer and Telecommunications Engineering WollongongNSW2522 Australia

It is challenging to localize a moving target using a multistatic system with unknown time delay (TD) and Doppler shift (DS) measurement offsets caused by asynchronization between transmitters and receivers. To address this issue, we develop a low complexity iterative message passing localization method using TD and DS measurements. In particular, to handle the unknown offsets, we formulate a joint estimation problem of target position, velocity and unknown offsets based on the maximum likelihood (ML) principle, which is nonlinear and nonconvex. Then, by transforming the objective function and relevant constraints into a product of multiple local functions, a message passing algorithm is developed. To handle the complex local functions in message updates, belief propagation (BP) and variational message passing (VMP) are combined. Comparisons with state-of-the-art methods demonstrate that, with much lower complexity, the proposed method delivers superior performance. © 2022 IEEE.

关键词： Iterative methods

来源：评论

学校读者我要写书评

暂无评论

Real-Time Disease Classification in Coffee Plants with Coral Developer Board Mini

Real-Time Disease Classification in Coffee Plants with Coral...

引用

2024 IEEE International Conference on Agrosystem engineering, Technology and Applications, AGRETA 2024

作者： Warsi, Mohammad Hamzah Nugroho, Hermawan Almurib, Haider Abbas Siang, Tay Fei Eswaran, Sivaraman University of Nottingham Malaysia Department of Electrical and Electronic Engineering Semenyih Malaysia Swinburne University of Technology Faculty of Engineering Computing and Science Sarawak Campus Sarawak Malaysia Curtin University Malaysia Department of Electrical and Computer Engineering Miri Malaysia

ISBN: (纸本)9798350373660

Coffee significantly contributes to the socioeconomic development of numerous countries. However, coffee production faces several challenges, including the detrimental effects of coffee leaf diseases, which impact crop yield and quality. Traditional methods for diagnosing coffee leaf diseases are time-consuming and prone to human error. This paper presents an automated system for coffee leaf disease classification with edge computing using the Coral Developer Board Mini, featuring Google's Edge TPU. The system leverages advanced convolutional neural network (CNN) models, including MobileNetV2, VGG-16, VGG19, and DenseNet121, to achieve high accuracy in disease identification. The MobileNetV2 model achieved the highest training accuracy of 98.9%. Inference time comparisons revealed that the Coral Developer Board Mini, with an inference time of 10 milliseconds, outperformed a traditional GPU-based workstation, which had an inference time of 40 milliseconds. Results highlight the efficiency and suitability of the Coral Developer Board Mini for real-time, on-device machine learning applications in agriculture. © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Review of Techniques for Integrating Security in Software Development Lifecycle

引用

computers, Materials & Continua 2025年第1期82卷 139-172页

作者： Hassan Saeed Imran Shafi Jamil Ahmad Adnan Ahmed Khan Tahir Khurshaid Imran Ashraf College of Electrical and Mechanical Engineering National University of Sciences and Technology(NUST)Rawalpindi46604Pakistan Department of Computer Science Abasyn University Islamabad CampusIslamabad45510Pakistan Military College of Signals National University of Sciences and Technology(NUST)Rawalpindi46600Pakistan Department of Electrical Engineering Yeungnam UniversityGyeongsan38541Republic of Korea Department of Information and Communication Engineering Yeungnam UniversityGyeongsan38541Republic of Korea

Software-related security aspects are a growing and legitimate concern,especially with 5G data available just at our *** conduct research in this field,periodic comparative analysis is needed with the new techniques coming up *** purpose of this study is to review the recent developments in the field of security integration in the software development lifecycle(SDLC)by analyzing the articles published in the last two decades and to propose a way *** review follows Kitchenham’s review *** review has been divided into three main stages including planning,execution,and *** the selected 100 articles,it becomes evident that need of a collaborative approach is necessary for addressing critical software security risks(CSSRs)through effective risk management/estimation *** risks using a numeric scale enables a comprehensive understanding of their severity,facilitating focused resource allocation and mitigation *** a comprehensive understanding of potential vulnerabilities and proactive mitigation efforts facilitated by protection poker,organizations can prioritize resources effectively to ensure the successful outcome of projects and initiatives in today’s dynamic threat *** review reveals that threat analysis and security testing are needed to develop automated tools for the *** estimation of effort required to prioritize potential security risks is a big challenge in software *** accuracy of effort estimation can be further improved by exploring new techniques,particularly those involving deep *** is also imperative to validate these effort estimation methods to ensure all potential security threats are *** challenge is selecting the right model for each specific security *** achieve a comprehensive evaluation,researchers should use well-known benchmark checklists.

关键词： Software development lifecycle systematic literature review critical software security risks national institute of standards and technology DevSecOps open web application security project McGraw’s touch points

来源：评论

学校读者我要写书评

暂无评论

Triplet Loss-less Center Loss Sampling Strategies in Facial Expression Recognition Scenarios 57

Triplet Loss-less Center Loss Sampling Strategies in Facial ...

引用

57th Annual Conference on Information sciences and Systems, CISS 2023

作者： Rajoli, Hossein Lotfi, Fatemeh Atyabi, Adham Afghah, Fatemeh Clemson University Holcombe Department of Electrical and Computer Engineering ClemsonSC United States University of Colorado Colorado Springs Department of Computer Science Colorado Springs United States

ISBN: (纸本)9781665451819

Facial expressions convey massive information and play a crucial role in emotional expression. Deep neural network (DNN) accompanied by deep metric learning (DML) techniques boost the discriminative ability of the model in facial expression recognition (FER) applications. DNN, equipped with only classification loss functions such as Cross-Entropy cannot compact intra-class feature variation or separate inter-class feature distance as well as when it gets fortified by a DML supporting loss item. The triplet center loss (TCL) function is applied on all dimensions of the sample's embedding in the embedding space. In our work, we developed three strategies: fully-synthesized, semi-synthesized, and prediction-based negative sample selection strategies. To achieve better results, we introduce a selective attention module that provides a combination of pixel-wise and element-wise attention coefficients using high-semantic deep features of input samples. We evaluated the proposed method on the RAF-DB, a highly imbalanced dataset. The experimental results reveal significant improvements in comparison to the baseline for all three negative sample selection strategies. © 2023 IEEE.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

Model Uncertainty Estimation Based on a Dual-Module Neural Network Architecture 23

Model Uncertainty Estimation Based on a Dual-Module Neural N...

引用

23rd International Conference on Machine Learning and Cybernetics, ICMLC 2024

作者： Guo, Weiqing Zhang, Tianlun Peng, Dunlu Wang, Debby D. School of Science and Technology Hong Kong Metropolitan Universit Hong Kong Hong Kong School of Optical-Electrical and Computer Engineering University of Shanghai for Science and Technology Shanghai China

ISBN: (纸本)9798331528041

In recent years, deep learning has achieved remarkable success in various fields, including biology and medicine. However, the interpretability and robustness still face challenges, as erroneous predictions can lead to serious consequences in critical applications. This paper introduces a new framework for estimating model uncertainty that can enhance the decision-making process. Compared to existing uncertainty estimation methods like Bayesian neural networks and Monte Carlo dropout, this paper provides a more direct way to model uncertainty by directly estimating the output variance through a dual-module neural network architecture. Without the need for multiple forward passes or complex Bayesian inference, it can reduce computational intensity. In specific, this framework consists of a prediction module for input processing and an uncertainty module for estimating prediction uncertainty. We demonstrate the efficiency of our method through applications in medical image segmentation and protein-ligand binding affinity prediction. By incorporating model uncertainty estimation, the proposed framework not only improves the accuracy of deep learning models, but also significantly reduces the computational burden which is especially beneficial for processing large datasets. © 2024 IEEE.

关键词： Prediction models

来源：评论

学校读者我要写书评

暂无评论

ALIEN-YOLO: Low-Light Object Detection with Edge Extraction and Convolutional Attention 10

ALIEN-YOLO: Low-Light Object Detection with Edge Extraction ...

引用

10th International Conference on computer and Communications, ICCC 2024

作者： Yang, Xinpeng Wang, Yongxiong School of Optical-Electrical and Computer Engineering University of Shanghai for Science and Technology Department of Automation Shanghai China

In the field of security monitoring, the detection of targets in low-light nighttime images is essential for identifying potential security threats. However, this task is challenged by factors such as increased noise, loss of detail, and reduced contrast, which often result in the suboptimal performance of existing detection algorithms under these conditions. To address these challenges, this study introduces an enhanced model based on YOLOv8, termed ALIEN-YOLO. This model incorporates the Adaptive Edge Extraction Low-light Image Enhancement Network (ALIEN), which is specifically designed to mitigate noise and enhance contrast. Additionally, it integrates the Head Depthwise Convolution Attention (HDCA) module to improve the detection of small and occluded objects, and the C2F-SH module, which is optimized with a Shrinkage mechanism to enhance robustness and accuracy through denoising. Experimental evaluations conducted on the ExDark and DARK FACE datasets demonstrated that the improved model achieved a mean Average Precision (mAP) of 81.2% and 54.3%, respectively, indicating a significant enhancement over the baseline YOLOv8. Notably, despite the integration of multiple modules, the model maintains a real-time processing capability of 60 frames per second. These findings underscore the effectiveness of the proposed ALIEN-YOLO model in enhancing target detection performance in low-light conditions, thereby underscoring its potential value for practical applications. © 2024 IEEE.

关键词： attention mechanism edge extraction image enhancement low-light target detection YOLOv8

来源：评论

学校读者我要写书评

暂无评论

Spoofing Face Detection Using Novel Edge-Net Autoencoder for Security

引用

Intelligent Automation & Soft Computing 2023年第3期35卷 2773-2787页

作者： Amal H.Alharbi S.Karthick K.Venkatachalam Mohamed Abouhawwash Doaa Sami Khafaga Department of Computer Sciences College of Computer and Information SciencesPrincess Nourah bint Abdulrahman UniversityRiyadh 11671Saudi Arabia Department of Electrical Electronics and Communication Engineering GITAM School of TechnologyGITAM Deemed to be UniversityBengaluru CampusKarnataka560065India Department of Computer Science and Engineering CHRIST(Deemed to be University)Bangalore560074India Department of Mathematics Faculty of ScienceMansoura UniversityMansoura 35516Egypt Department of Computational Mathematics Scienceand Engineering(CMSE)Michigan State UniversityEast LansingMI48824USA

Recent security applications in mobile technologies and computer sys-tems use face recognition for high-end *** numerous security tech-niques,face recognition is considered a high-security *** fuse and carry out face identiﬁcation as an access authority into these ***,face identiﬁcation authentication is sensitive to attacks with a 2-D photo image or captured video to access the system as an authorized *** the existing spooﬁng detection algorithm,there was some loss in the recreation of *** research proposes an unobtrusive technique to detect face spooﬁng attacks that apply a single frame of the sequenced set of frames to overcome the above-said *** research offers a novel Edge-Net autoencoder to select convoluted and dominant features of the input diffused ***,this pro-posed method is tested with the Cross-ethnicity Face Anti-spooﬁng(CASIA),Fetal alcohol spectrum disorders(FASD)*** database has three models of attacks:distorted photographs in printed form,photographs with removed eyes portion,and video *** images are taken with three different quality cameras:low,average,and high-quality real and spoofed *** extensive experimental study was performed with CASIA-FASD,3 Diagnostic Machine Aid-Digital(DMAD)dataset that proved higher results when compared to existing algorithms.

关键词： Image processing edge detection edge net auto-encoder face authentication digital security

来源：评论

学校读者我要写书评

暂无评论

Accelerating Deep Learning in Reconstructive Spectroscopy Using Synthetic Data

Accelerating Deep Learning in Reconstructive Spectroscopy Us...

引用

2023 Conference on Lasers and Electro-Optics, CLEO 2023

作者： Li, Pengyu Yaras, Can Sarwar, Tuba Ku, Pei-Cheng Qu, Qing Electrical Engineering and Computer Science University of Michigan 1301 Beal Ave Ann ArborMI48105 United States

A deep learning prototype for a reconstructive spectrometer was developed using an experimentally determined spectrometer response and synthetic data generated from the response matrix. Benchmarking with synthetic and... 详细信息

ISBN: (纸本)9781957171258

关键词： Spectrometers

来源：评论

学校读者我要写书评

暂无评论

Deep Moore-Penrose Inverse Network with Refinement Strategy for One-class Classification

Deep Moore-Penrose Inverse Network with Refinement Strategy ...

引用

2023 IEEE International Conference on Systems, Man, and Cybernetics, SMC 2023

作者： Gao, Junna Kong, Dehui Yin, Baocai Lin, Weisi Zhang, Wandong Beijing Institute of Artificial Intelligence Beijing University of Technology Beijing China School of Computer Science and Engineering Nanyang Technological University Singapore Western University Department of Electrical and Computer Engineering London Canada

ISBN: (纸本)9798350337020

Multilayer least-square (LS)-based one-class classification networks (MLS-OCNs) have gained great attention for the purpose of identifying anomalies and outliers. However, many MLS-OCNs encounter the issue of loosely connected feature coding because they use two separate mechanisms for feature encoding and final pattern recognition. This paper proposes a solution to this problem by introducing a multilayer algorithm called deep Moore-Penrose inverse network with refinement (DMPINR). In particular, DMPINR employs an end-to-end learning process based on the Moore-Penrose inverse (MPI) to identify optimal latent space and classify objects simultaneously. To enhance the robustness of representations, the DMPINR technique pulls back the residual error from the output layer to the hidden layers sequentially, recalculating the parameters of these hidden layers using MPI. The experimental results on ten popular OCC datasets demonstrate that the proposed approach outperforms many existing MLS-OCNs in G-Mean and F1 scores. © 2023 IEEE.

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：