检索结果-内蒙古大学图书馆

8th International Conference of Yearly Reports on Informatics, Mathematics, and engineering, ICYRIME 2023

作者： Fiani, Francesca Puglisi, Adriano Napoli, Christian Department of Computer Control and Management Engineering Sapienza University of Rome Roma00185 Italy Institute for Systems Analysis and Computer Science Italian National Research Council Roma00185 Italy Department of Computational Intelligence Czestochowa University of Technology Czestochowa42-201 Poland

Modern neural networks models for computer vision are trained on millions of images. The idea is that models are able to increase generalization when the dataset contains well diversified images, e.g. with varied illumination and environmental conditions of the same objects. Generalization is particularly relevant in object detection, especially for what concerns the cross-depiction problem. In this work we explore the use of Neural Style Transfer as a novel technique to morph the original data, with the aim to enhance model generalization. To verify the effect on performances for object detection models, we selected the Faster R-CNN model to be applied on the Pascal VOC 2012 dataset. A number of tests were performed through style variations on images and by tuning Neural Style Transfer parameters to maintain the content of the original images. The experiments showed promising results, which effectively provide a foundation for future studies on cross-depiction via Neural Style Transfer. © 2023 Copyright for this paper by its authors.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Can microplastic contamination affect the wing morphology and wingbeat frequency of Aedes aegypti (Diptera: Culicidae) mosquitoes?

引用

Environmental science and Pollution Research 2024年第49期31卷 59289-59301页

作者： de Nadai, Barbara Lepretti Moura, Lidia Castro, Gleyson B. Silva, Kamila Jessie Sammarro Maletzke, André Gustavo Corbi, Juliano José Batista, Gustavo E. A. P. A. Machado, Renato Bobsin Computational Entomology Laboratory Department of Engineering and Exact Sciences Western Paraná State University Foz Do Iguaçu Brazil Environmental Engineering and Sanitation Leonardo da Vinci University Center Indaial Brazil Laboratory of Ecotoxicology and Applied Ecology Department of Hydraulic Engineering and Sanitation São Carlos School of Engineering University of Sao Paulo Sao Carlos Brazil Environmental Biophotonics Laboratory São Carlos Institute of Physics University of Sao Paulo Sao Carlos Brazil School of Computer Science and Engineering University of New South Wales Sydney Australia

Microplastics (MPs) are increasingly widespread in the environment, which raises questions about their potential effects at different biological levels. It is essential to assess the impacts on biodiversity, and it is also crucial to understand whether the presence of MPs can interfere with the biological traits of species of relevance in public health. Considering that the life-history traits of mosquitoes, such as size and the wingbeat frequency (WBF), are related to its vector competence, here, we study the effects of 106 particles L−1 (as expected concentration of MPs on the environment, using the polyethylene type) on WBF, as well as wing morphology, testing the Culicidae species found across all continents, Aedes aegypti, as an indicator. Results show that larvae survival and development were not affected by the tested concentration of MP. Geometric morphometrics showed some asymmetry in female mosquito wings, which were also smaller for individuals reared in MP suspension. As for WBF, results did not indicate any significant differences between females. Male mosquitoes, however, showed alterations in WBF and wing morphology, suggesting possible sex-specific reactions to microplastic exposure. Also, the combination of morphological parameters analyzed as covariates (wing centroid size and body weight) did not significantly affect WBF for both female and male mosquitoes. Overall, this study shows an inaugural investigation of the effects of MP on wing size and WBF on Ae. aegypti, shedding light on these parameters tested for a current pollution issue and its impact on a virus vector. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024.

关键词： Invertebrates

来源：评论

学校读者我要写书评

暂无评论

ADAPTIVE RANK ALLOCATION: SPEEDING UP MODERN TRANSFORMERS WITH RANA ADAPTERS

arXiv

引用

arXiv 2025年

作者： Garcia, Roberto Liu, Jerry Sorvisto, Daniel Eyuboglu, Sabri Institute of Computational and Mathematical Engineering Stanford University United States Department of Computer Science Stanford University United States

Large Language Models (LLMs) are computationally intensive, particularly during inference. Neuron-adaptive techniques, which selectively activate neurons in Multi-Layer Perceptron (MLP) layers, offer some speedups but suffer from limitations in modern Transformers. These include reliance on sparse activations, incompatibility with attention layers, and the use of costly neuron masking techniques. To address these issues, we propose the Adaptive Rank Allocation framework and introduce the Rank and Neuron Allocator (RaNA) adapter. RaNA adapters leverage rank adapters, which operate on linear layers by applying both low-rank matrix decompositions and adaptive masking to efficiently allocate compute without depending on activation sparsity. This enables RaNA to be generally applied to MLPs and linear components of attention modules, while eliminating the need for expensive maskers found in neuron-adaptive methods. Notably, when compared to neuron adapters, RaNA improves perplexity by up to 7 points and increases accuracy by up to 8 percentage-points when reducing FLOPs by ∼44% in state-of-the-art Transformer architectures. These results position RaNA as a robust solution for improving inference efficiency in modern Transformer architectures. © 2025, CC BY.

关键词： Neurons

来源：评论

学校读者我要写书评

暂无评论

Writer Identification from Offline Handwriting Images in Urdu Script with Dense-Net: A Deep Learning Approach 14

Writer Identification from Offline Handwriting Images in Urd...

引用

14th International Conference on Computing Communication and Networking Technologies, ICCCNT 2023

作者： Nabi, Syed Tufael Singh, Paramjeet Kumar, Munish Maharaja Ranjit Singh Punjab Technical University Department of Computer Science and Engineering College of Engineering and Technology GZS Campus Punjab Bathinda151001 India Maharaja Ranjit Singh Punjab Technical University Department of Computational Sciences Punjab Bathinda151001 India

ISBN: (纸本)9798350335095

Handwriting serves as a valuable tool in various organizational processes, such as recruitment, interviewing, team-building, counseling, and career planning. Even identical twins with identical genes exhibit distinct handwriting styles. Leveraging this uniqueness, systems that can classify handwriting can determine the identity, age, handedness, nationality, and other details of the writer, making it an effective and reliable indicator of personality and behavior. Writer identification systems, for instance, can identify writers based on their distinct handwriting patterns. Manual extraction of handwriting features from a large dataset was once nearly impossible, but thanks to deep learning algorithms, it is now feasible to automatically extract features from documents and accurately predict the author's identity. However, there is a dearth of such systems for Urdu script, as evidenced by the literature survey. Therefore, in this paper, we propose a CNN-based model, Dense-Net, as a writer identification system trained and tested on an offline Urdu handwritten dataset contributed by 418 Urdu writers. Our model achieved an impressive overall accuracy of 99.20%, outperforming all existing methods for writer identification systems. © 2023 IEEE.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Graph Neural Networks for Optimizing Reliability in Vehicle-to-Infrastructure Communication in Autonomous Systems

Graph Neural Networks for Optimizing Reliability in Vehicle-...

引用

International Conference on Communication and Electronics Systems (ICCES)

作者： G Neelamegam J. Revathy M. Kannukkiniyal M. Devendran J. Akshya Mani Deepak Choudhry Department of Computer Science and Engineering VSB College of Engineering and Technology Coimbatore Department of Artificial Intelligence and Data Science Christ the King Engineering College Coimbatore Department of Computer Science and Engineering Kongu Engineering College Erode Department of Computer Science and Engineering Hindustan institute of technology Coimbatore Department of Computational Intelligence SRM Institute of Science and Technology Chennai Department of Computing Technologies SRM Institute of Science and Technology Chennai

ISBN: (数字)9798350377972

ISBN: (纸本)9798350377989

High reliance on autonomous systems necessitates efficient and reliable data exchange through Vehicle-to-Infrastructure (V2I) communication to have appropriate and stable network performance in dynamic vehicular environments. With heuristic approaches, Dijkstra's algorithm, and Deep Q-Networks (DQN), several methods are proposed for optimizing latency reduction, packet delivery rate improvement, and power efficiency in the direction of V2I communications. However, these approaches suffer from important challenges: low adaptability of the network to real-time dynamics, suboptimal resource usage, and difficulties in scalability towards complex large-scale networks. In this scenario, we propose a framework that incorporates a GNN inside autonomous systems for V2I communication to optimize performance. The GNN will process real-time graph features, including signal strength, packet loss, and mobility speed, as it dynamically changes in adaptation to the evolution of network behaviour and other functions. This work shows the superiority of the proposed GNN model for the efficiency of experimentally conducted simulations compared to the current models; it achieved latency reduced by up to 28%, 5-10% PDR, and 20% energy efficiency across the various scenarios with urban, rural, highway, and suburban networks. The results validate robustness and scalability for practical implementations by the proposed approach in real-world autonomous systems. Further development is possible with this framework: for example, the use of multi-agent reinforcement learning techniques within cooperative decision-making and the enhancement through blockchain-based security mechanisms to ensure the integrity of the data.

关键词： Road transportation Adaptation models Vehicle-to-infrastructure Scalability Heuristic algorithms Reinforcement learning Graph neural networks Real-time systems Energy efficiency Vehicle dynamics

来源：评论

学校读者我要写书评

暂无评论

Writer Identification: Deep Learning with ResNet50 for Offline Urdu Handwritten Documents 7

Writer Identification: Deep Learning with ResNet50 for Offli...

引用

7th International Conference on Image Information Processing, ICIIP 2023

作者： Nabi, Syed Tufael Singh, Paramjeet Kumar, Munish Maharaja Ranjit Singh Punjab Technical University College of Engineering And Technology Department of Computer Science And Engineering GZS Campus Punjab Bathinda151001 India Maharaja Ranjit Singh Punjab Technical University Department of Computational Sciences Punjab Bathinda151001 India

ISBN: (纸本)9798350371406

The field of writer identification, a crucial aspect of document analysis, has undergone a significant evolution through the incorporation of deep learning models. This paper presents an innovative writer identification system, meticulously designed around the ResNet50 deep learning architecture, with a specific emphasis on its adaptability for processing offline Urdu handwritten documents. The system underwent rigorous training and thorough testing using an Urdu handwritten dataset that encompasses contributions from 418 Urdu writers. Our ResNet50-based model demonstrates an exceptional level of accuracy, achieving an impressive identification rate of 99.26%. Within this paper, we delve into a comprehensive examination of the model's architecture, intricate feature extraction methods, and the subtle intricacies of its training methodologies, all of which collectively contribute to this outstanding performance. Beyond pushing the boundaries of current writer identification capabilities, the application of deep learning, particularly the ResNet 50 model, to Urdu handwritten documents holds significant promise across various domains, including historical document preservation, forensic analysis, and robust authentication. This paper not only introduces a state-of-the-art writer identification system but also underscores the broader impact of deep learning in the field of document analysis. The remarkable accuracy demonstrated by our ResNet50-based model highlights the transformative potential of deep learning, enhancing the precision and effectiveness of writer identification tasks. Furthermore, it illuminates new avenues for research and practical applications, spanning diverse linguistic and cultural contexts. © 2023 IEEE.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Mosquito Type Identification using Convolution Neural Network 3

Mosquito Type Identification using Convolution Neural Networ...

引用

3rd International Conference on Smart Electronics and Communication, ICOSEC 2022

作者： Kumar, V Senthil Prasad, Veduri Veera Kunaisa, Khallik Swain, Mahendra Pratap Anandaram, Harishchander Kumar, Ashok Malla Reddy Institute of Technology and Science Department of Computer Science and Engineering Telangana Secunderabad500100 India Aditya College of Engineering Department of Computer Science and Engineering Andhra Pradesh Surampalem India Hkbk College of Engineering Department of Computer Science and Engineering Karnataka Bengaluru560045 India Department of Pharmaceutical Sciences and Technology Jharkhand Mesra Ranchi835215 India Amrita Vishwa Vidyapeetham Centre for Excellence in Computational Engineering and Networking Tamil Nadu Coimbatore641112 India Banasthali Vidyapith Department of Computer Science Rajasthan Banasthali304022 India

ISBN: (数字)9781665497640

ISBN: (纸本)9781665497640

Mosquitoes are insects which are most commonly found in places with high water content. There are various types of mosquitos and each type has its own destructive characteristics. In order to completely eradicate mosquitoes from a particular place, people must know what type of mosquito it is. But this is not a task that can be done with naked eyes. To help this situation, this study develops a Deep Learning (DL) model which can predict the type of mosquito with great accuracy. For this purpose, a dataset of images of the Aedes and culex species of mosquito is collected from Kaggle. Following that, this dataset is split into three parts: training, validation, and testing. Then, pixel and image scaling is used to preprocess this dataset. A DL model is produced using the Convolutional Neural Network (CNN) technique. The model is then trained using the preprocessed dataset. The results of the training are also calculated and documented. The results are then put into a graph for easier examination. The highest accuracy of the model is produced during training and it is 99.2%. As for the loss values, the lowest loss value is 0.8% which is so small for a classification algorithm. Next, tests are conducted on the trained and validated models. A confusion matrix is then used to examine the model performance. The model is then assessed using the metrics such as accuracy, True Positive Rate (TPR), True Negative Rate (TNR), False Positive Rate (FPR), False Negative Rate (FNR), precision, and F1 score. These parameters' final values were also determined to be satisfactory. © 2022 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Distributed Bearing-Only Formation Control for Heterogeneous Nonlinear Multi-Robot Systems

引用

IFAC-PapersOnLine 2023年第2期56卷 3447-3452页

作者： Kefan Wu Junyan Hu Zhengtao Ding Farshad Arvin Department of Electrical and Electronic Engineering The University of Manchester UK Department of Computer Science University College London UK Swarm & Computational Intelligence Laboratory (SwaCIL) Department of Computer Science Durham University UK

This paper addresses the bearing-only formation tracking problem for heterogeneous nonlinear multi-robot systems. In contrast to position and distance-based formation algorithms, the robots can only measure the bearing information from their neighbors to achieve cooperation while the state information is unavailable. This characteristic is able to be implemented in the hardware to reduce the requirements of the sensors. We construct a compensation function in the proposed controller to eliminate the effect of the unknown nonlinear terms in the system. This compensation function is also based on bearing measurements, which guarantees that the overall controller is bearing-only. The stability of the proposed formation tracking strategy can be ensured by Lyapunov techniques. Moreover, we analyze the performance of the protocol for moving leaders, where the formation tracking error can be restricted in a bounded set. Finally, the simulation results are presented to validate the feasibility of the proposed algorithm for both fixed and moving leaders.

关键词： Bearing measurements formation tracking nonlinear control multi-robot systems

来源：评论

学校读者我要写书评

暂无评论

Decision Making in Problem of Irrigation Based on Z-Information 16th

Decision Making in Problem of Irrigation Based on Z-Informat...

引用

16th International Conference on Application of Fuzzy Systems, Soft Computing and Artificial Intelligence Tools, ICAFS 2023

作者： Zeinalova, Lala M. Muravyova, Elena Bogdanova, Diana Department of Computer Engineering Azerbaijan State University of Oil and Industry 20 Azadlig Avenue Baku1010 Azerbaijan Department of Automated Technological and Information Systems Ufa State Petroleum Technological University Sterlitamak Russia Department of Computational Mathematics and Cybernetics Ufa University of Science and Technology Ufa Russia

ISBN: (纸本)9783031762826

In this paper an intelligent irrigation system is considered. Irrigation regimes for agricultural crops provide the determination of scientifically based norms of irrigation, depending on the soil, physiological, climatic and technical characteristics of the irrigated array. In modern irrigation systems, the parameters of irrigation regimes are calculated similarly to traditional methods. To solve the problem of efficient irrigation, it is necessary to take into account the weak formalizability, which makes it difficult to use accurate mathematical models to describe the control system. The solution to this problem would be partially possible if there was a large number of qualified experts who, taking into account the linguistic nature of the variables, are able to determine the optimal irrigation rates, using their experience without constructing any functional dependencies. Due to the absence of such reference experts in intelligent irrigation planning systems, mathematical models based on Soft Computing should come forward. Our model is based on Z-information, which has such an important qualitative attribute as reliability. The concept of Z-number relates to the issue of reliability of information, especially in the realms of decision analysis. Z-number valuations take into account the uncertаinty of the experts` opinion in the estimаtion of the options. We apply an interpolation approach for Z-rules for the solution of problem of approximate reasoning. A practical example of evaporation is given to show the applicability of approach in irrigation process. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

关键词： Fuzzy logic

来源：评论

学校读者我要写书评

暂无评论

Educational Digital Twin: Tackling Complexity in Educational Big Data

Educational Digital Twin: Tackling Complexity in Educational...

引用

IEEE International Conference on Big Data

作者： Luwen Huang Karen E. Willcox Department of Computer Science University of Texas Austin Oden Institute for Computational Engineering & Sciences University of Texas Austin

ISBN: (数字)9798350362480

ISBN: (纸本)9798350362497

"Everything is bigger in Texas" and this includes the big data challenges of the state’s educational system. But just as big is the opportunity for digital twin technologies to improve decision-making in education, with potential to improve student outcomes. This paper formulates an Educational Digital Twin, a novel approach to understanding, modeling, and analyzing educational data to address the complexities inherent in student pathways. By "student pathways", we refer to the joint facets of student behavior—evolving dynamically over time and encompassing not only course enrollments but also behavioral and academic factors. Unlike traditional approaches that rely on static analyses of historical data, the Educational Digital Twin applies the digital twin paradigm to model these pathways as a living construct that dynamically evolves alongside the physical world. At the heart of the Educational Digital Twin is a knowledge graph that organizes data into a semantic structure. By employing graph-theoretic methods, we manipulate this structure to derive multi-granular insights that inform decision-making. Our evaluation demonstrates how the Educational Digital Twin not only facilitates the dynamic integration of new data but also scales efficiently to handle complex datasets covering millions of students across hundreds of demographic and academic dimensions.

关键词： Analytical models Decision making Semantics Static analysis Big Data Data models Digital twins Complexity theory Stakeholders Security

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：