检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Li, Zhiyuan Zhou, Yanhui Wei, Hao Ge, Chenyang Mian, Ajmal Institute of Artificial Intelligence and Robotics Xi’an Jiaotong University Xi’an710049 China Department of Computer Science and Software Engineering The University of Western Australia Crawley PerthWA6009 Australia

Diffusion-based extreme image compression methods have achieved impressive performance at extremely low bitrates. However, constrained by the iterative denoising process that starts from pure noise, these methods are limited in both fidelity and efficiency. To address these two issues, we present Relay Residual Diffusion Extreme Image Compression (RDEIC), which leverages compressed feature initialization and residual diffusion. Specifically, we first use the compressed latent features of the image with added noise, instead of pure noise, as the starting point to eliminate the unnecessary initial stages of the denoising process. Second, we directly derive a novel residual diffusion equation from Stable Diffusion’s original diffusion equation that reconstructs the raw image by iteratively removing the added noise and the residual between the compressed and target latent features. In this way, we effectively combine the efficiency of residual diffusion with the powerful generative capability of Stable Diffusion. Third, we propose a fixed-step fine-tuning strategy to eliminate the discrepancy between the training and inference phases, thereby further improving the reconstruction quality. Extensive experiments demonstrate that the proposed RDEIC achieves state-of-the-art visual quality and outperforms existing diffusion-based extreme image compression methods in both fidelity and efficiency. The source code and pre-trained models will be released. Copyright © 2024, The Authors. All rights reserved.

关键词： Iterative methods

来源：评论

学校读者我要写书评

暂无评论

Improving Predictive Navigation Through the Optimization of Counterfactual Track Evaluation

Improving Predictive Navigation Through the Optimization of ...

引用

IEEE Symposium on Position Location and Navigation (PLANS)

作者： Alexander Stringer Geoffrey Dolinger Timothy Sharp Debra Hogue Joseph Karch Lesya Borowska Justin G. Metcalf US Air Force's Oklahoma City Air Logistics Complex and the 76 Software Engineering Group School of Electrical and Computer Engineering and the Advanced Radar Research Center at the University of Oklahoma Department of Computer Science University of Oklahoma

Modern navigation applications are confronting the challenge of path identification and prediction. Because of the inherent noisiness of measurement techniques, positional data often contains errors. Over time, when attempting to predict future positions, these errors can compound, causing drift in the target's path. Filtering or regression techniques can use periodic high-fidelity measurements, like those from GPS sensors, to correct positional data and reduce historic errors. These techniques have been much less effective in specialized environments where GPS is intermittent or denied. At the same time, the development of semi- and fully-autonomous systems has increased the need for accurate predictive navigation. To address this problem, previous methods were proposed for repairing tracking data using causality-aware machine learning (ML). It combined a long short-term memory (LSTM) network that predicted target paths with the non-dominated sorting genetic algorithm II (NSGA-II) which identified and evaluated counterfactual paths. Here te authors expand on that work by improving the simulated data used to train and test the system, expanding the GA by adding additional first principles-based objective functions, improving upon the LSTM implementation, and utilizing Extended Kalman Filter (EKF) pre-processing. System testing is conducted on Matlab-generated navigation scenarios, and results are compared to both EKF correction and spline interpolation. The proposed counterfactual track repair (CTR) tool produces paths with less repeatability than traditional approaches. However, it is shown to consistently generate more realistic path corrections, lower error in predictions based on its corrected paths, and is highly configurable, demonstrating the value and utility of this approach.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Identification of Data Breaches from Public Forums 1

引用

14th International Conference on Innovative Security Solutions for Information Technology and Communications, SecITC 2021

作者： Adnan, Md. Akhtaruzzaman Younus, Atika Kawser, Md. Harun Al Adhikary, Natasha Habib, Ahsan Haque, Rakib Ul Department of Computer Science and Engineering University of Asia Pacific Dhaka1205 Bangladesh Indetechs Software Limited House: 31 Road: 20 Block: K Banani Dhaka1212 Bangladesh School of Computer Science & Technology University of Chinese Academy of Sciences Shijingshan Beijing100049 China

ISBN: (数字)9783031175107

ISBN: (纸本)9783031175091

Adversaries initiate their cyberattacks towards different entities such as healthcare or business institutes, and a successful attack causes data breaches. They publish their success stories in public forums for ranking purposes. The victim entities can be informed early about the data breach event if these forums are analyzed properly. Though few studies already focused on this sector, their data sets and codes are not public. Most importantly, the sources of their data sets do not exist today, which makes their novelty unclear and unreliable. To address and handle the above concerns, this study reinvestigates this domain with Machine Learning, Ensemble Learning, and Deep Learning. A web crawler is developed for downloading the dataset from the public forum of Nulled website. Feature extraction is done using TF-IDF and GloVe. Performance analysis showed that SVM achieved at most 90.80% accuracy with linear kernel. Implementations are published with a GitHub link. © 2022, Springer Nature Switzerland AG.

关键词： Web crawler

来源：评论

学校读者我要写书评

暂无评论

An End-to-End Test Case Prioritization Framework using Optimized Machine Learning Models

An End-to-End Test Case Prioritization Framework using Optim...

引用

IEEE International Conference on software Testing Verification and Validation Workshop, ICSTW

作者： Md Asif Khan Akramul Azim Ramiro Liscano Kevin Smith Yee-Kang Chang Gkerta Seferi Qasim Tauseef Department of Electrical Computer and Software Engineering Ontario Tech University Oshawa Ontario L1G 0C5 International Business Machines Corporation (IBM)

ISBN: (数字)9798350344790

ISBN: (纸本)9798350344806

Regression testing in software development is challenging due to the large number of test cases and continuous integration (CI) practices. Recently, test case prioritization (TCP) using machine learning (ML) has been shown to efficiently execute regression tests. This study introduces an automated, endto-end, self-contained ML-based framework, TCP-Tune, tailored exclusively for TCP. The framework utilizes open-source version control system data to combine code-change-related features with test execution results. This integration allows the automated optimization of hyperparameters across different ML models to improve the TCP. The framework also effectively visualizes and utilizes multiple evaluation metrics to evaluate the performance of the model over several builds. Unlike existing implementations, which rely on various frameworks, TCP-Tune enables the effortless incorporation of features from multiple sources and fine-tuned models, thereby providing optimum test prioritization in the ever-changing field of software development. Our approach has helped to provide efficient TCP through experimental assessments of a real-life, large-scale CI system.

关键词： software testing Radio frequency Measurement Machine learning Production Continuous integration Market research

来源：评论

学校读者我要写书评

暂无评论

Enhance Training Objectives for Image Captioning with Decomposed Sequence-level Metric

Enhance Training Objectives for Image Captioning with Decomp...

引用

International Joint Conference on Neural Networks (IJCNN)

作者： Yehuan Wang Jiefeng Long Jian Hu Lin Shang Department of Computer Science and Technology State Key Laboratory for Novel Software Technology Nanjing University Nanjing China School of Mechanical Engineering Nanjing University of Science and Technology Nanjing China

ISBN: (数字)9798350359312

ISBN: (纸本)9798350359329

Image captioning aims to generate fluent and accurate descriptions for images. To evaluate the quality of captions, various metrics have been proposed. However, current metrics only assess captions at sequence-level, which overwhelms the distinctions between each token. Thus, existing objectives tend to treat each token equally, assigning them with identical weights in loss functions. Intuitively, key words in a caption carry the primary information and contribute more than other words to sequence-level metrics. They should be distinguished and weighted more during training. In this work, we propose to explicitly measure each word and guide the model to focus more on key words in captions. Firstly, we devise token-level CIDEr (CIDEr-T) as a new metric to quantify the importance of each word, by decomposing the sequence-level CIDEr into token-level granularity. CIDEr-T maintains consistency with CIDEr and shows the distinctions between tokens. Thus, we engage CIDEr-T scores of each token as their unique weights in the raw loss functions, which can bridge the gap between training and evaluation.

关键词： Measurement Training Accuracy Neural networks

来源：评论

学校读者我要写书评

暂无评论

Report from Dagstuhl Seminar 21442: Ensuring the Reliability and Robustness of Database Management Systems

引用

Datenbank-Spektrum 2022年第3期22卷 261-263页

作者： Rigger, Manuel Böhm, Alexander Christakis, Maria Lo, Eric Department of Computer Science National University of Singapore 13 Computing Drive Singapore Singapore Google Germany GmbH ABC-Str. 19 Hamburg Germany Max Planck Institute for Software Systems Paul-Ehrlich Strasse G 26 Kaiserslautern Germany Department of Computer Science and Engineering The Chinese University of Hong Kong Ho Sin-Hang Eng. Building Hong Kong

In November 2021, Dagstuhl seminar 21442 was bringing together researchers and practitioners from various domains such as of databases, automatic testing, and formal methods to build a common ground and to explore possibilities for systematically improving the state of the art in database management system engineering. The outcome of the seminar was a joint understanding of the specific intricacies of building stateful system software, as well as the identification of several areas of future work. In particular, we believe that database system engineering can both be significantly improved by adopting additional verification techniques and testing tools, and can provide important feedback and additional challenges (e.g. related to state management) to neighboring domains. © 2022, Gesellschaft für Informatik e.V. and Springer-Verlag GmbH Germany, part of Springer Nature.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Hand Written Digit Recognition using Multilayer Deep Convolutional Neural Network

Hand Written Digit Recognition using Multilayer Deep Convolu...

引用

Smart Technologies, Communication and Robotics (STCR)

作者： S Deepa G D Praveenkumar Vivek Duraivelu R Suguna P Sathishkumar N B Priyananthan Department of Computer Technology - PG Kongu Engineering College Perundurai Erode Tamilnadu Department of computer science VET Institute of Arts and Science(Co-Education)College Erode Tamilnadu Department of Computer Science and Engineering B V Raju Institute of Technology Narsapur Telangana Department of Computer science and Engineering Bannari Amman Institute of Technology Sathyamangalam Final M.Sc(Software Systems) Kongu Engineering College Perundurai Erode Tamilnadu

In machine learning, handwritten digit recognition is usually seen as a multi-class classification problems In this approach, the ten possible digits (0-9) are treated as individual classes, and the goal is to train a classifier that can accurately identify them. However, it’s not unusual for a single classifier to have varying levels of success when applied to different datasets, even after being trained using a standard learning algorithm. This indicates that while a given learning algorithm may be effective at training strong classifiers on certain datasets, it may result in weaker classifiers for others. Furthermore, it’s possible for a classifier to exhibit varying levels of performance on multiple test datasets, especially considering that different writers may produce highly diverse image samples of the same numbers. To address this issue, the advancement of ensemble learning methodologies will be critical, as they have the potential to improve overall prediction accuracy and offer more consistent performance across different datasets.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Knowing Users with Mild Intellectual Impairments in UX Design Processes 23

Knowing Users with Mild Intellectual Impairments in UX Desig...

引用

Proceedings of the XXIII International Conference on Human computer Interaction

作者： Barbara Corsetti Raquel HerváS Ballesteros Lourdes Moreno Computer Science and Engineering Department Universidad Carlos III de Madrid Spain Departamento de Ingeniería del Software e Inteligencia Artificial Universidad Complutense de Madrid Spain

ISBN: (纸本)9798400707902

Living with mild intellectual disabilities leads many people to struggle against several daily difficulties. For instance, a user with mild intellectual impairments is often not able to complete basic tasks independently (e.g., moving around the city, managing the home, shopping) and, additionally, he/she always depends on another person like a caregiver or a legal guardian. Starting from these hypotheses, the motivation behind this work is to design an interactive system for supporting users with mild intellectual impairments by improving their autonomy and, consequentially, the quality of their lives. In order to understand the needs and problems of this user group, this paper presents the first stage of a UX design process based on the Double Diamond model that was carried out using the Empathy Map tool to visualize the information gathered in the interviews with 15 users with mild intellectual impairments and three of their caregivers. By applying this approach, we reach a full understanding of the users and their environment which, as conclusions of this work, allows us to set the most appropriate decisions about the future interactive system design and development. The Empathy Map tool has turned out to be a valid tool for use with the specific target of people with intellectual disabilities, however, some tool extensions could be considered to be carried out in the future.

关键词： Design Double Diamond Empathy Map Intellectual Impairments UX User Interview

来源：评论

学校读者我要写书评

暂无评论

Multiple Feature Selection Frameworks Based on Evolutionary Computing and Ensemble Learning for software Defect Prediction

Multiple Feature Selection Frameworks Based on Evolutionary ...

引用

International Conference on Computational Intelligence and Networks (CINE)

作者： Ruchika Malhotra Anmol Budhiraja Abhinav Kumar Singh Ishani Ghoshal Shweta Meena Department of Software Engineering Delhi Technological University Delhi India Department of Computer Science and Engineering Delhi Technological University Delhi India

ISBN: (纸本)9781665464666

software defects may cause severe crashes in the system, leading to the software's high maintenance costs. Early identification of these defects would lead to high-quality software, thus saving time and money. This study proposes five feature selection approaches based on evolutionary computing algorithms, each coupled with a majority voting ensemble for software defect prediction. The objective is to improve the existing process by targeting the metric selection stage. The study was conducted on thirty open-source defect datasets. The proposed feature selection techniques were applied on a within-project defect prediction model and a heterogeneous defect prediction model. The Friedman and the Wilcoxon Signed-rank test concluded that the proposed techniques were promising and generated results comparable to some other state-of-the-art feature selection methodologies.

关键词： Measurement Statistical analysis Computational modeling software algorithms Evolutionary computation Predictive models Maintenance engineering

来源：评论

学校读者我要写书评

暂无评论

An Intelligent Method for Detecting Jute Pests Using Deep Convolutional Neural Networks

An Intelligent Method for Detecting Jute Pests Using Deep Co...

引用

IoT Based Control Networks and Intelligent Systems (ICICNIS), International Conference on

作者： S. Vignesh S. Prabhu R. Sampathkumar M. Thillai Rani J. Rejina Parvin G. Uma Maheswari R. Saravanakumar Department of Computer Applications Karpagam Academy of Higher Education Coimbatore Tamil Nadu India Department of Mechatronics Engineering KPR Institute of Engineering and Technology Coimbatore Tamil Nadu India Department of Electronics and Communication Engineering Sri Krishna College of Technology Coimbatore Tamil Nadu India Department of Electronics and Communication Engineering Sri Krishna College of Engineering and Technology Coimbatore Tamil Nadu India Department of Computer Science and Engineering RMK College of Engineering and Technology Chennai Tamil Nadu India Iconix Software Solution Tirunelveli Tamil Nadu India

ISBN: (数字)9798331518097

ISBN: (纸本)9798331518103

Pest attacks pose a serious threat to the production of jute and other significant crops. Jute farmers often utilize their visual sense and hands-on expertise to distinguish between multiple diseases that seem to be identical. The intelligent model we built for the identification of jute pests was based on a deep convolutional neural network (DCNN). This practical difficulty led to the development of this concept. The proposed DCNN model can automatically identify jute pest concerns with high speed and accuracy based on image recognition. In addition, images of the four most common jute bugs are included in an organized image graphics collection. Our model yields a DCNN accuracy is 96.72% for the four most significant jute pest kinds. Further proof of the model's effectiveness is provided by the accuracy, recall, F1-score, and confusion matrix metrics.

关键词： Performance evaluation Visualization Accuracy Operating systems Production Pest control Feature extraction software Convolutional neural networks Overfitting

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：