检索结果-内蒙古大学图书馆

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Gourav Datta Zeyu Liu Anni Li Peter A. Beerel Dept. of Electrical Computer & Systems Engineering Case Western Reserve University Cleveland USA Ming Hsieh Dept. of Electrical and Computer Engineering University of Southern California Los Angeles USA

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Spiking Neural Networks (SNNs) have emerged as a popular spatio-temporal computing paradigm for complex vision tasks. Recently proposed SNN training algorithms have significantly reduced the number of time steps (down to 1) for improved latency and energy efficiency, however, they target only convolutional neural networks (CNN). These algorithms, when applied to the recently spotlighted vision transformers (ViT), either require a large number of time steps or fail to converge. Based on the analysis of the histograms of the ANN and SNN activation maps, we hypothesize that each ViT block has a different sensitivity to the number of time steps. We propose a novel training framework that dynamically allocates the number of time steps to each ViT module depending on a trainable score assigned to each timestep. In particular, we generate a scalar binary time step mask that filters spikes emitted by each neuron in a leaky-integrate-and-fire (LIF) layer. The resulting SNNs have high activation sparsity and require only accumulate operations (AC), except for the input embedding layer, in contrast to expensive multiply-and-accumulates (MAC) needed in traditional ViTs. This yields significant improvements in energy efficiency. We evaluate our training framework and resulting SNNs on image recognition tasks including CIFAR10, CIFAR100, and ImageNet with different ViT architectures. We obtain a test accuracy of 95.97% with 4.97 time steps with direct encoding on CIFAR10.

关键词： Training computer vision Sensitivity Neurons Signal processing algorithms Spiking neural networks Transformers Energy efficiency Convolutional neural networks Speech processing

来源：评论

学校读者我要写书评

暂无评论

Augmented Karuta: Interactive Playing Cards on the Floor for Learning Local Culture

Augmented Karuta: Interactive Playing Cards on the Floor for...

引用

IEEE International Conference on Artificial Intelligence and Virtual Reality (AIVR)

作者： Nobuyuki Umezu Hideki Bando Dept. of Computer and Information Sciences Ibaraki University Hitachi Japan Dept. of Mechanical Systems Engineering Ibaraki University Hitachi Japan

ISBN: (数字)9798331521578

ISBN: (纸本)9798331521585

This paper introduces Augmented Karuta, an interactive floor projection experience inspired by traditional Japanese playing cards, known as ‘Karuta’. Designed to engage users and stimulate interest in local cultures, history, and traditions, the system utilizes three projectors. Two projectors project 46 “grabbing” cards onto the floor screen of 4.1 x 3.7 (m), while the third displays a corresponding “reading” card on a wall screen. Each matching pair of cards shares a poem or sentence that begins with one of the Japanese phonograms. During a session, users actively engage by stepping onto the projected ‘grabbing’ card that matches the ‘reading’ card displayed on the wall, effectively blending physical movement with cognitive skills to find the correct match. By integrating a two-dimensional LiDAR sensor, the system tracks the user's position on the interactive floor. This tracking is vital for the proposed system to immediately respond to the user's actions, ensuring an immersive and seamless interaction with the digital karutas on the floor. We conducted evaluation experiments at two museums, gathering positive feedback from 57 participants. Future work will focus on quantitatively evaluating the system's impact on learning effectiveness.

关键词： Laser radar Tracking Virtual reality Learning (artificial intelligence) Museums History Floors

来源：评论

学校读者我要写书评

暂无评论

Dual Auction Peer-to-Peer control Scheme for Empowering Prosumers Participation in the Local Energy Markets

Dual Auction Peer-to-Peer Control Scheme for Empowering Pros...

引用

International Conference on Computing, Mathematics and engineering Technologies (iCoMET)

作者： Muhammad Ikram Salman Ahmed Safdar Nawaz Khan Marwat Dept. of Computer Systems Engineering University of Engineering and Technology Peshawar Pakistan

The emerging trend of distributed and renewable energy sources encourages energy trading in the electricity market. Various forms of energy markets are evolving for successful energy trading. The local energy market is one of them in which various participants are actively coordinating for optimal utilization of power. Such type of local energy markets provides individual prosumers to trade their surplus energy in the neighbourhood in peer-to-peer topology. The complexity of such type of local energy market is increasing in terms of structure, control, coordination, and cost estimation. For the cooperative behaviour of the local energy markets, a dual auction coordination control scheme is important to encourage more prosumers for energy trading. We proposed a novel scheme that performs dual auctions for prosumers in a decentralized fashion. We assessed dual auction matching by coordinating prosumers and consumers to sell and purchase energy in the local energy market. The aggregator performed balancing in the proposed dual auction scheme as an autonomous entity. The results were simulated in MATLAB and JADE for the complex prosumer peer-to-peer network. Our proposed scheme shows each peer dual auction matching improves from 0.95 to 2.03 for optimal hierarchal energy trading in energy markets. For peer participation, the participation ratio increased from 0.684 to 1.74 and optimal aggregator coordination from 9.18% to 19% in the local energy markets.

关键词： Renewable energy sources Costs Estimation Consensus control Market research Electricity supply industry Peer-to-peer computing

来源：评论

学校读者我要写书评

暂无评论

Network-based Intrusion Detection Datasets: A Survey 23

Network-based Intrusion Detection Datasets: A Survey

引用

23rd International Arab Conference on Information Technology, ACIT 2022

作者： Ahmed, Lubna Ali Hassan Hamad, Yahia Abdalla Mohamed Abdalla, Ahmed Abdallah Mohamed Ali Sudan University of Science and Technology Dept. of Electronics Engineering Khartoum Sudan Sudan University of Science and Technology Dept. of Computer Networks and Systems Khartoum Sudan

ISBN: (纸本)9798350320244

The increasing reliance on computers and the internet has increased the need for protection against attacks and security threats. A network-based intrusion detection system (NIDS) is one of the promising research fields that could highly contribute to cyber-Attacks detection and mitigation. Datasets are necessary for NIDS training and evaluation. This paper provides a comprehensive survey of publicly available NIDS datasets. The study covers the most used NIDS datasets, each dataset is briefly described along with its main properties and shortcomings. Moreover, the paper highlights the most important dataset characteristics and provides a clear datasets comparison accordingly. Finally, the paper discusses observations and provides some recommendations to help researchers identify future work. © 2022 IEEE.

关键词： Network security

来源：评论

学校读者我要写书评

暂无评论

LightNN: An Optimized Deep Learning Framework for Enhanced Fruit Classification

LightNN: An Optimized Deep Learning Framework for Enhanced F...

引用

Computing, Applications and systems (COMPAS), IEEE International Conference on

作者： Maharin Afroj Israt Jahan Md Polash Islam Md. Saifur Rahman Dept. of Computer Science and Engineering Bangladesh University of Business and Technology Dhaka Bangladesh Dept. of Computer and Systems Sciences Stockholm University Stockholm Sweden

ISBN: (数字)9798331529765

ISBN: (纸本)9798331529772

The crucial field of fruit classification in computer vision and deep learning brings about breakthroughs that have broad applications in the retail and agricultural industries. By automating the identification and sorting of various fruit types, these technologies enhance efficiency, accuracy, and scalability, benefiting processes from farm production to market distribution. The integration of sophisticated algorithms and neural networks not only improves the speed and reliability of fruit classification but also offers the potential for real-time quality control and inventory management, revolutionizing traditional methods and paving the way for more innovative, data-driven approaches in these sectors. This study introduces LightNN, a novel and effective paradigm with a lightweight architecture that tackles the complex problems related to computational efficiency. Combining a small classifier with a simplified feature extractor, LightNN achieves the ideal balance between simplicity and accuracy. The feature extractor is capable of extracting hierarchical features from RGB images by using convolutional layers with ReLU activations and max-pooling procedures. The flattened output is then processed by the classifier, which consists of fully connected layers with ReLU activations and dropout regularisation. Remarkably, LightNN, with just 298,835 parameters, manages to get a respectable 96.37% test accuracy in fruit categorization tasks. The results highlight LightNN’s promise as an effective and efficient fruit classifying method, particularly in situations with constrained computational resources.

关键词： Deep learning Training Visualization Accuracy Computational modeling Scalability computer architecture Feature extraction Computational efficiency Sorting

来源：评论

学校读者我要写书评

暂无评论

Improving ASTRO's Perception Capability for a more Natural Follow-Me Task 6

Improving ASTRO's Perception Capability for a more Natural F...

引用

6th Experiment at International Conference, exp.at 2023

作者： Ferreira, Bruno Sorrentino, Alessandra Menezes, Paulo Batista, Jorge Dias, Jorge Fiorini, Laura Cavallo, Filipo University of Coimbra Institute of Systems and Robotics Dept. of Electrical and Computer Engineering Coimbra Portugal Khalifa University of Science and Technology Khalifa University Center for Autonomous Robotic Systems Abu Dhabi United Arab Emirates University of Florence Dept. of Industrial Engineering Florence Italy BioRobotics Institute Scuola Superiore Sant'Anna Pisa Italy

ISBN: (纸本)9798350314403

Social assistive robots usually encompass a great compromise between the advanced perception models that one can use and their computing capabilities. The ideal approaches are always oriented towards low power consumption while maintaining a higher order of responsiveness to the surroundings. Therefore, we present in this paper, an improvement of the follow-me system on ASTRO. In detail, we propose the use of Mediapipe SDK for human detection and tracking, when ASTRO is meant to accompany someone that is walking. A study on the new velocity and distance profiles that the robot keeps from the individuals is presented and we have also evaluated how it affects their perception of being safe. The presented results show that our new approach allows the system to achieve real-time performance by becoming ≈ 9.1× faster, smoothing, and keeping a more natural distance from the user. © 2023 IEEE.

关键词： Human robot interaction

来源：评论

学校读者我要写书评

暂无评论

Remaining Useful Life Prediction of Multi-Component System Clustering with Time-Varying Kernel Density

Remaining Useful Life Prediction of Multi-Component System C...

引用

2023 China Automation Congress, CAC 2023

作者： Li, Ke Fan, Xingle Shi, Hui Zhang, Xiaohong Sun, Chaoli Wang, Jianan Taiyuan University of Science and Technology Shanxi Key Laboratory of Advanced Control and Equipment Intelligence Dept. of Electronic and Information Engineering Taiyuan China Taiyuan University of Science and Technology Shanxi Key Laboratory of Advanced Control and Equipment Intelligence Dept. of Economics and Management Taiyuan China Taiyuan University of Science and Technology Shanxi Key Laboratory of Advanced Control and Equipment Intelligence Dept. of Computer Science and Technology Taiyuan China

ISBN: (纸本)9798350303759

Aiming at the uncertainty of the degradation procedure of a complex multi-component system in a time-varying environment and the influence of the random dependence between different components, a remaining useful life prediction model for time-varying adaptive kernel density estimation based on the dependence between multiple components is established. First, the random dependence between multi-component systems is clustered. Considering that different components in the same cluster are affected by the random dependence, the components between different clusters do not affect each other. Judge the dependence characteristics between components, and use the dependence characteristics of different components to conduct clustering degradation modelling. Secondly, the parameters of the system change with time. Considering that monitoring data near the current time has a greater impact on the remaining useful life prediction compared with historical data, data in the vicinity of the current time is accorded a more significant weight, and a time-varying kernel density estimation prediction method is designed, in which the nearest-neighbour adaptive window-width selection method is used. Finally, give an example to demonstrate the effectiveness by the proposed method. © 2023 IEEE.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

Features of Space Vector Pulse Width Modulation in 5-phases Cascaded High Voltage Converters 5

Features of Space Vector Pulse Width Modulation in 5-phases ...

引用

5th IEEE KhPI Week on Advanced Technology, KhPIWeek 2024

作者： Busher, Victor Chornyi, Oleksii Tytiuk, Valerii Glazeva, Oksana Kuznetsov, Vitaliy Dept. of Electrical Engineering and Electronics National University "Odessa Maritime Academy" Odessa Ukraine Institute of Electromechanics Energy Saving and Automatic Control Systems Kremenchuk Mykhailo Ostrohradskyi National University Kremenchuk Ukraine Kryvyi Rih National University Kryvyi Rih Ukraine Dept. of the electrical engineering Ukrainian State University of Science and Technologies Dnipro Ukraine

ISBN: (纸本)9798350384499

The purpose of this work is to analyze the capabilities of space-vector pulse-width modulation when controlling a 5-phase high-voltage cascade frequency converter in normal mode and several types of emergency situations. Taking into account the large number of controlled power transistors, an asynchronous modulation method is proposed, based on calculating the amplitude values of individual phases at each time based on the duty cycle of space vectors, which allows the control system to use independent switching in each phase. In emergency modes, when individual H-bridges are damaged, it is proposed to use the balanced space vectors method, which ensures the symmetry of the electromagnetic field in the event of a short circuit of both individual cascades and one or two phases of the inverter. Methodology. We use a mathematical and geometric interpretation of variable amplitude space vectors when projecting voltages onto a symmetrical 5-phase load. Results. Analysis of the 1st harmonic of phase voltages showed that the use of the space vector modulation method in the normal operating mode of a 5-phase converter makes it possible to increase the utilization factor of DC voltage sources by 23.1% (instead of 15.6% for a 3-phase converter). If individual H-bridges fail and the voltages in the load phases are balanced, the utilization factor may decrease, but in any case it will remain greater than 1. In this case, the higher harmonics, despite significant distortions in the shape of the phase voltages, are mutually compensated in the electromagnetic field of the load, as evidenced by absence of pulsations and shocks on the graphs of transient processes of the electromagnetic torque of the engine. Balancing of the electromagnetic field in the load is maintained even if one or two phases of the inverter are short-circuited. This significantly increases the survivability of the electric drive, which is especially important in critical mechanisms and technological processes,

关键词： Electric motors

来源：评论

学校读者我要写书评

暂无评论

GFT: Gradient Focal Transformer

arXiv

引用

arXiv 2025年

作者： Kriuk, Boris Gill, Simranjit Kaur Aslam, Shoaib Fakhrutdinov, Amir Hong Kong University of Science and Technology Dept. of Computer Science & Engineering Hong Kong University of Westminster Dept. of Computer Science & Engineering United Kingdom University of Engineering & Technology Lahore Dept. of Mechanical Mechatronics & Control Eng Pakistan Shanghai Jiao Tong University Dept. of Mechanical Engineering China

Fine-Grained Image Classification (FGIC) remains a complex task in computer vision, as it requires models to distinguish between categories with subtle localized visual differences. Well-studied CNN-based models, while strong in local feature extraction, often fail to capture the global context required for fine-grained recognition, while more recent ViT-backboned models address FGIC with attention-driven mechanisms but lack the ability to adaptively focus on truly discriminative regions. TransFG and other ViT-based extensions introduced part-aware token selection to enhance attention localization, yet they still struggle with computational efficiency, attention region selection flexibility, and detail-focus narrative in complex environments. This paper introduces GFT (Gradient Focal Transformer), a new ViT-derived framework created for FGIC tasks. GFT integrates the Gradient Attention Learning Alignment (GALA) mechanism to dynamically prioritize class-discriminative features by analyzing attention gradient flow. Coupled with a Progressive Patch Selection (PPS) strategy, the model progressively filters out less informative regions, reducing computational overhead while enhancing sensitivity to fine details. GFT achieves SOTA accuracy on FGVC Aircraft, Food-101, and COCO datasets with 93M parameters, outperforming ViT-based advanced FGIC models in efficiency. By bridging global context and localized detail extraction, GFT sets a new benchmark in fine-grained recognition, offering interpretable solutions for real-world deployment scenarios. Copyright © 2025, The Authors. All rights reserved.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Comparative Evaluation of Machine Learning-Based controller Area Network Intrusion Detection systems 14

Comparative Evaluation of Machine Learning-Based Controller ...

引用

14th International Conference on Information and Communication systems, ICICS 2023

作者： Al-Jarrah, Omar Y. Ramadan, Omar A. Alsaify, Baha Alkhushayni, Suboh Jordan University of Science and Technology Faculty of Computer and Information Technology Dept. Network Engineering and Security P.O. Box 3030 Irbid22110 Jordan Jordan University of Science and Technology Faculty of Engineering Dept. Mechanical Engineering P.O. Box 3030 Irbid22110 Jordan Yarmouk University Dept. Information Systems Irbid Jordan

ISBN: (纸本)9798350307870

The controller area network (CAN) protocol is widely used in vehicle networks. However, it lacks essential security features like confidentiality and authentication. To enhance vehicle security, researchers have proposed and developed various methods, including intrusion detection systems (IDSs). These IDSs play a key role in early detection of potential cyberattacks, safeguarding vehicles, passengers, and road users. Machine learning (ML) algorithms are increasingly employed in CAN IDSs but they face challenges due to diverse experimental settings, datasets, and metrics used for evaluation, impeding progress and comparison of different approaches in this field. This study aims to benchmark well-recognized ML algorithms, including gaussianNB, k-nearest neighbors (KNN), decision trees (DT), random forest (RF), long short-term memory (LSTM), and convolutional neural networks (CNN), for intrusion detection in CAN. We evaluate the performance of these algorithms using seven evaluation metrics on the real ORNL automotive dynamometer (ROAD) dataset, a state-of-the-art benchmark in this field. Experimental results demonstrate that KNN, DT, RF, LSTM, and CNN can detect various types of attacks (e.g., Fuzzing, targeted ID, and masquerade attacks) with high accuracy. However, the computational and time efficiency of tree-based classifiers, namely DT and RF, makes them a potentially appealing choice for real-time intrusion detection tasks in CAN. © 2023 IEEE.

关键词： Intrusion detection

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：