检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Lakatos, Róbert Pollner, Péter Hajdu, András Joó, Tamás Data-Driven Health Division of National Laboratory for Health Security Health Services Management Training Centre Semmelweis University Hungary Department of Data Science and Visualization Faculty of Informatics University of Debrecen Hungary Doctoral School of Informatics University of Debrecen Hungary Neumann Technology Platform Neumann Nonprofit Ltd

The development of generative large language models (G-LLM) opened up new opportunities for the development of new types of knowledge-based systems similar to ChatGPT, Bing, or Gemini. Fine-tuning (FN) and Retrieval-Augmented Generation (RAG) are the techniques that can be used to implement domain adaptation for the development of G-LLM-based knowledge systems. In our study, using ROUGE, BLEU, METEOR scores, and cosine similarity, we compare and examine the performance of RAG and FN for the GPT-J-6B, OPT-6.7B, LlaMA, LlaMA-2 language models. Based on measurements shown on different datasets, we demonstrate that RAG-based constructions are more efficient than models produced with FN. We point out that connecting RAG and FN is not trivial, because connecting FN models with RAG can cause a decrease in performance. Furthermore, we outline a simple RAG-based architecture which, on average, outperforms the FN models by 16% in terms of the ROGUE score, 15% in the case of the BLEU score, and 53% based on the cosine similarity. This shows the significant advantage of RAG over FN in terms of hallucination, which is not offset by the fact that the average 8% better METEOR score of FN models indicates greater creativity compared to RAG. © 2024, CC BY.

关键词： Knowledge based systems

来源：评论

学校读者我要写书评

暂无评论

The Classification of Timor Weaving Using Artificial Neural Network

The Classification of Timor Weaving Using Artificial Neural ...

引用

International Conference on informatics and Computing (ICIC)

作者： Mendarissan Aritonang Adi Setiawan Krismiyati Theophilus Wellem Dept of Computer Science Faculty of Information Technology Satya Wacana Christian University Salatiga Indonesia Dept. Mathematics and Data Science Faculty of Science and Mathematics Satya Wacana Christian University Salatiga Indonesia Dept. Informatics & Computer Engineering Education Faculty of Information Technology Satya Wacana Christian University Salatiga Indonesia Department of Informatics Satya Wacana Christian University Salatiga Indonesia

ISBN: (数字)9798331517601

ISBN: (纸本)9798331517618

The number of woven fabric on the island of Timor is very large and varied, making it difficult to distinguish between types and origins. Many woven fabric motifs appear similar but represent different types. Therefore, this classification was used to perform pattern recognition of woven fabric in the system. The process allowed the system’s ability to be tested by applying the pattern recognition algorithm to woven images. In this context, the feature extraction method used was gray level co-occurrence matrix (GLCM), while the method used for classification was artificial neural network (ANN). The results showed an accuracy of $87.5 \%$ in the system. Finally, GUI system was developed, enabling tests to be performed on woven images for the identification of Timor Weaving image patterns.

关键词： Training Accuracy Social networking (online) Neurons Artificial neural networks Feature extraction Weaving Fabrics Pattern recognition Web sites

来源：评论

学校读者我要写书评

暂无评论

Nice Cross-gramian, the Generalized Symmetry and Determination of Controller Structure for Multivariable Switched Systems

Nice Cross-gramian, the Generalized Symmetry and Determinati...

引用

2022 International Conference on Electrical, Computer, Communications and Mechatronics Engineering, ICECCME 2022

作者： Tahavori, Maryamsadat Shaker, Hamid Reza Dashti, Rahman Technical University of Denmark DTU Engineering Technology Ballerup Denmark Center for Energy Informatics University of Southern Denmark Odense Denmark Persian Gulf University Faculty of Intelligent Systems Engineering & Data Science Bushehr*** Iran

ISBN: (数字)9781665470957

ISBN: (纸本)9781665470957

In this paper, the determination of a control struc-ture for discrete-time MIMO switched systems is addressed. In the first step, for discrete-time switched systems, the concept of nice cross-gramians is proposed and investigation is done on whether the suggested nice cross-gramian exists. It has been demonstrated that the nice cross-gramian can be achieved as a solution to a specific algebraic matrix equation, if it exists. While the proposed cross-gramian has a wide range of uses, including model reduction, in this paper, it has been utilized to form the Nice Hankel Interaction Index Array (NHIIA) for control structure selection. The control structure determination strategy which is proposed is one of the few that supports switched systems, takes advantage of gramian-based techniques, and is computationally more effective than its only counterpart. © 2022 IEEE.

关键词： Matrix algebra

来源：评论

学校读者我要写书评

暂无评论

DSLR-CNN: Efficient CNN Acceleration using Digit-Serial Left-to-Right Arithmetic

arXiv

引用

arXiv 2025年

作者： Nisar, Malik Zohaib Ibrahim, Muhammad Sohail Gorgin, Saeid Usman, Muhammad Lee, Jeong-A Department of Computer Engineering College of IT Convergence Chosun University Gwangju61452 Korea Republic of Faculty of Informatics and Data Science University of Regensburg Regensburg93053 Germany Tehran33535-111 Iran

Digit-serial arithmetic has emerged as a viable approach for designing hardware accelerators, reducing interconnections, area utilization, and power consumption. However, conventional methods suffer from performance and latency issues. To address these challenges, we propose an accelerator design using left-to-right (LR) arithmetic, which performs computations in a most-significant digit first (MSDF) manner, enabling digit-level pipelining. This leads to substantial performance improvements and reduced latency. The processing engine is designed for convolutional neural networks (CNNs), which includes low-latency LR multipliers and adders for digit-level parallelism. The proposed DSLR-CNN is implemented in Verilog and synthesized with Synopsys design compiler using GSCL 45nm technology, the DSLR-CNN accelerator was evaluated on AlexNet, VGG-16, and ResNet-18 networks. Results show significant improvements across key performance evaluation metrics, including response time, peak performance, power consumption, operational intensity, area efficiency, and energy efficiency. The peak performance measured in GOPS of the proposed design is 4.37× to 569.11× higher than contemporary designs, and it achieved 3.58× to 44.75× higher peak energy efficiency (TOPS/W), outperforming conventional bit-serial designs. © 2025, CC BY.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Quadratic Sets and (3mod5)-Arcs in PG(r,5) 20th

Quadratic Sets and (3mod5)-Arcs in PG(r,5)

引用

20th EAI International Conference on Computer science and Education in Computer science, CSECS 2024

作者： Landjev, Ivan Kurz, Sascha Rousseva, Assia Bulgarian Academy of Sciences Institute of Mathematics and Informatics 8 Acad G. Bonchev str. Sofia1113 Bulgaria New Bulgarian University 21 Montevideo str. Sofia1618 Bulgaria Mathematisches Institut Universität Bayreuth Bayreuth95440 Germany Department of Data Science Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen91058 Germany Faculty of Mathematics and Informatics Sofia University 5 J. Bourchier blvd. Sofia1164 Bulgaria

To goal of this contribution is to provide a characterization of the (3mod5)-arcs in PG(r,5), r≥4. Such arcs are either lifted or quadratic and by a construction described by Kurz-Landjev-Pavese-Rousseva. &... 详细信息

ISBN: (纸本)9783031843112

关键词： (tmodq)-arcs divisible arcs quadrics

来源：评论

学校读者我要写书评

暂无评论

Inference-Based No-Learning Approach on Pre-Trained BERT Model Retrieval

Inference-Based No-Learning Approach on Pre-Trained BERT Mod...

引用

International Conference on Big data and Smart Computing (BIGCOMP)

作者： Huu-Long Pham Ryota Mibayashi Takehiro Yamamoto Makoto P. Kato Yusuke Yamamoto Yoshiyuki Shoji Hiroaki Ohshima Graduate School of Information Science University of Hyogo Kobe Hyogo Japan Faculty of Library Information and Media Science University of Tsukuba Tsukuba Ibaraki Japan School of Data Science Nagoya City University Nagoya Aichi Japan Faculty of Informatics Shizuoka University Hamamatsu Shizuoka Japan

In recent years, the practice of leveraging pre-trained machine learning models for specific tasks has gained traction. Instead of training models from the ground up, it is now common to fine-tune existing pre-trained models. However, users have the responsibility to select a pre-trained model that is suitable with their task-a challenge given the number of pre-trained models available. Conventionally, the suitability of a pre-trained model for a task is ascertained through fine-tuning, which is costly in term of time and computational resources. This research introduces an efficient retrieval method for BERT pre-trained models in document classification tasks. Our contributions are as follows: (i) We defined the problem of pre-trained model retrieval; (ii) We developed a benchmark dataset by fine-tuning and evaluating twenty distinct pre-trained BERT models across seventeen document classification tasks; (iii) We propose a method to retrieve suitable pre-trained BERT models without actual fine-tuning.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Digital Twin System for Smart Campuses

Digital Twin System for Smart Campuses

引用

IEEE International Conference on Control science and Systems Engineering (CCSSE)

作者： Huijie Wang Yuanyuan Cai Fabu Yue Hao Luo Sina Abdipoor Say Leng Goh School of Big Data BaoShan University BaoShan Yunnan China Department of Computer Science Universiti Putra Malaysia Malaysia Faculty of Computing and Informatics Universiti Malaysia Sabah Labuan Malaysia

ISBN: (数字)9798331517199

ISBN: (纸本)9798331517205

In an era where educational institutions are increasingly embracing technology to improve efficiency and enhance learning environments, the concept of smart campuses has emerged as a crucial development. The traditional approach of monitoring campus environments through video footage and manual inspections is becoming outdated as technology advances. To empower smarter campus management and enhance the educational experience, this study presents a systematic approach for smart campuses built on the foundation of digital twins. This technology creates a virtual replica of the physical campus, enabling real-time monitoring of various aspects like security, teaching, maintenance, environment, assets, and energy usage. We simulate real-world campus behavior by constructing this digital counterpart. The resulting digital twin image can be visualized through various user-friendly interfaces, including virtual reality and 3D models, empowering managers to make informed decisions for a more efficient and secure campus environment. Implementing a digital twin system in a smart campus setting demonstrates significant improvements in operational efficiency, security management, and environmental sustainability. By leveraging the power of digital twins, educational institutions can transition towards smarter, more responsive, and sustainable campus environments, ultimately enriching the educational experience for students and staff.

关键词： Uniform resource locators Three-dimensional displays databases Virtual reality Real-time systems Digital twins Maintenance Servers Monitoring Thermal stability

来源：评论

学校读者我要写书评

暂无评论

Comparative Analysis of Robotic Arm Efficiency: Evaluating PID and Fuzzy-PID Control using Printed Mechanical Structure and ESP8266 Integration

Comparative Analysis of Robotic Arm Efficiency: Evaluating P...

引用

International Conference on Cybernetics and Intelligent System (ICORIS)

作者： Padma Nyoman Crisnapati I Made Darma Susila I Komang Agus Ady Aryanto Jaturapith Krohkaew Ricky Aurelius Nurtanto Diaz I Made Suandana Astika Pande Department of Information Systems Faculty of Informatics and Computers Institute Technology and Business STIKOM Bali Denpasar Indonesia Department of Computer Systems Faculty of Informatics and Computers Institute Technology and Business STIKOM Bali Denpasar Indonesia Dept. of Big Data Management and Analytics Faculty of Science and Technology Rajamangala University of Technology Thanyaburi Thanyaburi Thailand

ISBN: (数字)9798331539603

ISBN: (纸本)9798331539610

This study investigates the comparative effectiveness of manual PID tuning versus Fuzzy-PID control in enhancing the performance of a robot arm with two degrees of freedom (2-DOF). The experimental approach involves conducting two scenarios: manual tuning of PID parameters in Scenario 1 and utilizing a fuzzy algorithm for PID parameter tuning in Scenario 2. The novelty of this research lies in the new 3D mechanical design of the 2-DOF robot arm and the utilization of the ESP8266 microcontroller for implementing Fuzzy-PID control. Performance evaluation is conducted using Root Mean Square Error (RMSE) and Root Mean Square Percentage Error (RMSPE) calculations. The results reveal that Fuzzy-PID control significantly reduces errors compared to manual PID tuning, with lower RMSE and RMSPE values indicating more precise and stable control. These findings underscore the potential of integrating fuzzy logic with PID control to enhance system adaptability and flexibility. Additionally, the ESP8266 is capable of handling the computations required for fuzzy-PID control while being cost-effective.

关键词： Fuzzy logic PI control Microcontrollers 2-DOF Manuals Manipulators Three-dimensional printing PD control Root mean square Tuning

来源：评论

学校读者我要写书评

暂无评论

SINK-FREE ORIENTATIONS: A LOCAL SAMPLER WITH APPLICATIONS

arXiv

引用

arXiv 2025年

作者： Anand, Konrad Freifeld, Graham Guo, Heng Wang, Chunyang Wang, Jiaheng School of Informatics University of Edinburgh Informatics Forum EdinburghEH8 9AB United Kingdom State Key Laboratory for Novel Software Technology New Cornerstone Science Laboratory Nanjing University 163 Xianlin Avenue Jiangsu Province Nanjing China Faculty of Informatics and Data Science University of Regensburg Bajuwarenstrasse 4 Regensburg93053 Germany

For sink-free orientations in graphs of minimum degree at least 3, we show that there is a deterministic approximate counting algorithm that runs in time O((n73/Ε72) log(n/Ε)), a near-linear time sampling algorithm, and a randomised approximate counting algorithm that runs in time O((n/Ε)2 log(n/Ε)), where n denotes the number of vertices of the input graph and 0 Copyright © 2025, The Authors. All rights reserved.

关键词： Graph algorithms

来源：评论

学校读者我要写书评

暂无评论

LHU-NET: A LIGHT HYBRID U-NET FOR COST-EFFICIENT, HIGH-PERFORMANCE VOLUMETRIC MEDICAL IMAGE SEGMENTATION

arXiv

引用

arXiv 2024年

作者： Sadegheih, Yousef Bozorgpour, Afshin Kumari, Pratibha Azad, Reza Merhof, Dorit Faculty of Informatics and Data Science University of Regensburg Regensburg93053 Germany Faculty of Electrical Engineering and Information Technology RWTH Aachen University Aachen52062 Germany Fraunhofer Institute for Digital Medicine MEVIS Bremen28359 Germany

The rise of Transformer architectures has revolutionized medical image segmentation, leading to hybrid models that combine Convolutional Neural Networks (CNNs) and Transformers for enhanced accuracy. However, these models often suffer from increased complexity and overlook the interplay between spatial and channel features, which is vital for segmentation precision. We introduce LHU-Net, a streamlined Hybrid U-Net for volumetric medical image segmentation, designed to first analyze spatial and then channel features for effective feature extraction. Tested on five benchmark datasets (Synapse, LA, Pancreas, ACDC, BRaTS 2018), LHU-Net demonstrated superior efficiency and accuracy, notably achieving a 92.66 Dice score on ACDC with 85% fewer parameters and a quarter of the computational demand compared to leading models. This performance, achieved without pre-training, extra data, or model ensembles, sets new benchmarks for computational efficiency and accuracy in segmentation, using under 11 million parameters. This achievement highlights that balancing computational efficiency with high accuracy in medical image segmentation is feasible. Our implementation of LHU-Net is freely accessible to the research community on GitHub. © 2024, CC BY.

关键词： Convolutional neural networks

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：