检索结果-内蒙古大学图书馆

Optimizing Vision Transformers: Unveiling 'Focus and Forget' for Enhanced computational efficiency

IEEE ACCESS 2025年 13卷 27908-27927页

作者： Latibari, Banafsheh Saber Homayoun, Houman Sasan, Avesta Univ Calif Davis Elect & Comp Engn Davis CA 95616 USA

Vision Transformers are renowned for their accuracy in computer vision tasks but are computationally and memory expensive, making them challenging to deploy on resource-constrained edge devices. In our research paper, we introduce a revolutionary approach to designing energy-aware dynamically prunable Vision Transformers for use in edge applications. Our solution denoted as Incremental Resolution Enhancing Transformer (IRET), works by the sequential sampling of the input image. However, in our case, the embedding size of input tokens is considerably smaller than prior-art solutions. This embedding is used in the first few layers of the IRET vision transformer until a reliable attention matrix is formed. Then the attention matrix is used to sample additional information using a learnable 2D lifting scheme only for important tokens and IRET drops the tokens receiving low attention scores. Hence, as the model pays more attention to a subset of tokens for its task, its focus and resolution also increase. This incremental attention-guided sampling of input and dropping of unattended tokens allow IRET to significantly prune its computation tree on demand. By controlling the threshold for dropping unattended tokens and increasing the focus of attended ones, we can train a model that dynamically trades off complexity for accuracy. Moreover, using early exiting our model is capable of doing anytime prediction. This is especially useful for real-word energy-sensitive edge devices, where accuracy and complexity could be dynamically traded based on factors such as battery life, reliability, etc.

关键词： Transformers computational modeling Computer vision Head Accuracy Predictive models Visualization Discrete wavelet transforms computational efficiency Spatial resolution deep learning pruning vision transformer

来源：评论

学校读者我要写书评

暂无评论

Development of a Novel Real-Time Weighted Surrogate Model for Wind Turbine-Based on DFIG: Enhancing computational efficiency in Power Systems With Error Bound

引用

IEEE ACCESS 2025年 13卷 50000-50016页

作者： Latif, Muhammad Ambreen, Hira Diyin, Tang Hassan, Farrukh Imran, Muhammad Abbasi, Saddam Hussain Imran, Muhammad Natl Univ Sci & Technol Coll Signals Dept Elect Engn Islamabad 44000 Pakistan Beihang Univ Aeronaut & Astronaut Sch Automat Sci & Elect Engn Dept Detect & Automat Engn Beijing 100191 Peoples R China Sunway Univ Fac Engn & Technol Dept Data Sci & Artificial Intelligence Darul Ehsan 47500 Selangor Malaysia Henan Univ Sch Comp & Informat Engn Kaifeng 475004 Henan Peoples R China

In the realm of modern wind turbine engineering, where precision is paramount for stability, control, and observability, this article introduces a groundbreaking method leveraging time-weighted Gramians. The focal point of this work is the reduction of model order in wind turbines featuring a double-fed induction generator with time-varying rotational speeds. Employing sophisticated state-space representations, we establish a comprehensive and systematic framework for analyzing wind turbine performance, ensuring adherence to stringent grid standards. A distinctive aspect of our proposed approach lies in the utilization of time-weighted Gramians and an innovative balanced realization technique to reduce the dimensionality of large state-space models effectively. Stability and reduced approximation errors are guaranteed by creating a lower-order system. A significant improvement is the availability of an a priori formula for error-bound, which allows for more efficient and faster computations. The use of time-weighted Gramians allowed for the application of this groundbreaking technique to time-sensitive systems in the real world, such as wind turbines. The optimization of models utilizing vast simulation data is what makes our methodology better than current methods. Wind turbines that allow for real-time adjustment of rotating speeds are part of this dataset. A lot of consideration is given to stability and error calculation in this study, which makes it innovative. The use of time-weighted Gramians in practical, real-time systems has greatly improved the accuracy and efficiency of modeling approaches.

关键词： Wind turbines Stators Mathematical models Doubly fed induction generators Rotors Power system stability Read only memory Reduced order systems computational modeling computational efficiency Time-weighted Gramians balance algorithm model reduction error-bound induction generator wind turbine

来源：评论

学校读者我要写书评

暂无评论

A Symmetric Projection Space and Adversarial Training Framework for Privacy-Preserving Machine Learning with Improved computational efficiency

引用

APPLIED SCIENCES-BASEL 2025年第6期15卷 3275-3275页

作者： Li, Qianqian Zhou, Shutian Zeng, Xiangrong Shi, Jiaqi Lin, Qianye Huang, Chenjia Yue, Yuchen Jiang, Yuyao Lv, Chunli China Agr Univ Beijing 100083 Peoples R China Beijing Foreign Studies Univ Beijing 100089 Peoples R China Univ Int Business & Econ Beijing 100029 Peoples R China China Univ Polit Sci & Law Beijing 102249 Peoples R China Peking Univ Beijing 100871 Peoples R China

This paper proposes a data security training framework based on symmetric projection space and adversarial training, aimed at addressing the issues of privacy leakage and computational efficiency encountered by current privacy protection technologies when processing sensitive data. By designing a new projection loss function and combining autoencoders with adversarial training, the proposed method effectively balances privacy protection and model utility. Experimental results show that, for financial time-series data tasks, the model using the projection loss achieves a precision of 0.95, recall of 0.91, and accuracy of 0.93, significantly outperforming the traditional cross-entropy loss. In image data tasks, the projection loss yields a precision of 0.93, recall of 0.90, accuracy of 0.91, and mAP@50 and mAP@75 of 0.91 and 0.90, respectively, demonstrating its strong advantage in complex tasks. Furthermore, experiments on different hardware platforms (Raspberry Pi, Jetson, and NVIDIA 3080 GPU) show that the proposed method performs well on low-computation devices and exhibits significant advantages on high-performance GPUs, particularly in terms of computational efficiency, demonstrating good scalability and efficiency. The experimental results validate the superiority of the proposed method in terms of data privacy protection and computational efficiency.

关键词： privacy protection data security adversarial training computational efficiency high-dimensional data compression

来源：评论

学校读者我要写书评

暂无评论

Enhanced computational efficiency and Accuracy of Discrete-Time PID-DPWM Equation

引用

IEEE ACCESS 2025年 13卷 62288-62305页

作者： Kim, Eungnam Choi, Jin-Young Korea Univ Dept Comp Sci Seoul 02841 South Korea Korea Univ Grad Sch Informat Secur Seoul 02841 South Korea

The technique of using the duty cycle output of a proportional integral derivative controller as the input for a count-based digital pulse width modulation module to process digital PWM signals is widely applied across various industrial fields, including automotive, power control, and motor control. This paper refers to this approach as the classic PID-DPWM equation. Classic PID-DPWM equations have been studied with the perspective that the raw-data measured input, and the integer-value duty value are integers. This study presents the improved PID-DPWM equations that start with the initial data scaling equation and perform integer-based computations and one floating-point operation. Consequently, the proposed equations become equivalent to the classic PID-PWM equations, enhancing computational efficiency and accuracy. The enhancement in computational efficiency was verified mathematically, and the improvement in accuracy was validated using performance index tests. A validation test of a closed-loop system in MATLAB/Simulink demonstrated that the improved standard PID-DPWM equation outperformed the classic standard PID-DPWM equation, achieving improvements in performance indices (ISE = 0.0001%, IAE = 0.27%, IE = 0.035%, ITSE = 0.012%, and ITAE = 2.85%). Furthermore, the improved difference PID-DPWM equation demonstrates further improvements over the classic difference PID-DPWM equation, achieving improvements in performance indices (ISE = 0.98%, IAE = 4.79%, IE = 15.87%, ITSE = 4.25%, and ITAE = 26.89%). Overall, the improved standard PID-DPWM equation exhibited enhanced performance in ISE, IAE, and ITAE, whereas the improved difference PID-DPWM equation exhibited superior performance in IE and ITSE.

关键词： Mathematical models Standards Accuracy computational efficiency Software Real-time systems Hardware Control systems Pulse width modulation Optimization Computation accuracy computation speed DPWM PID software control

来源：评论

学校读者我要写书评

暂无评论

Deep Q-Learning-Based Optimization of Path Planning and Control in Robotic Arms for High-Precision computational efficiency

引用

INTERNATIONAL JOURNAL OF ADVANCED COMPUTER SCIENCE AND APPLICATIONS 2025年第1期16卷 1199-1207页

作者： Li, Yuan Min, Byung-Won Liu, Haozhi Nanchong Vocat & Tech Coll Sch Intelligent Mfg Nanchong 637100 Sichuan Peoples R China Mokwon Univ Div Informat & Commun Convergence Engn Daejeon 35349 South Korea Nanchong Vocat & Tech Coll Div Continuing Educ Nanchong 637100 Sichuan Peoples R China

Optimizing path planning and control in robotic arms is a critical challenge in achieving high-precision and efficient operations in various industrial and research applications. This study proposes a novel approach leveraging deep Q-learning (DQL) to enhance robotic arm movements' computational efficiency and precision. The proposed framework effectively addresses key challenges such as collision avoidance, path smoothness, and dynamic control by integrating reinforcement learning techniques with advanced kinematic modelling. To validate the effectiveness of the proposed method, a simulated environment was developed using a 6-degree-of-freedom robotic arm, where the DQL model was trained and tested. Results demonstrated a significant performance improvement, achieving an average path optimization accuracy of 98.76% and reducing computational overhead by 22.4% compared to traditional optimization methods. Additionally, the proposed approach achieved real-time response capabilities, with an average decision-making latency of 0.45 seconds, ensuring its applicability in time-critical scenarios. This research highlights the potential of deep Q-learning in revolutionizing robotic arm control by combining precision and computational efficiency. The findings bridge gaps in robotic path planning and pave the way for future advancements in autonomous robotics and industrial automation. Further studies can explore the scalability of this approach to more complex and real-world environments, solidifying its relevance in emerging technological domains.

关键词： Optimization deep Q-learning path planning robotic arms precision computational efficiency kinematic

来源：评论

学校读者我要写书评

暂无评论

Resource Optimization in IoT Systems : A Hybrid AI-based Approach for Enhancing computational efficiency and Reducing Latency

Resource Optimization in IoT Systems : A Hybrid AI-based App...

引用

International Conference on ent Computing and Control Systems (ICICCS)

作者： Savitha. N M Jayaprakash Elavarasi. T Shivakumar. E K. C. Gayathri Moorthy Agoramoorthy Department of Computer Science and Engineering Saveetha School of Engineering Saveetha Institute of Medical and Technical Sciences Saveetha University and Technical Sciences Saveetha University Chennai Tamilnadu India

ISBN: (数字)9798331512088

ISBN: (纸本)9798331512095

With the rapid proliferation of IoT and smart devices, optimizing resource management in edge computing environments is critical to ensuring efficient computational performance and minimal latency. This study proposes a hybrid AI-based framework that integrates reinforcement learning with heuristic optimization techniques, including genetic algorithms, to dynamically predict and allocate resources based on real-time workload characteristics. The proposed model enhances computational efficiency by optimizing resource distribution, reducing latency, and improving system scalability. Comparative experiments indicate that the AI-driven approach achieves up to a 30% improvement in computational efficiency and scalability compared to conventional resource allocation methods. This adaptive framework provides a robust and scalable solution for real-time data processing in IoT networks, significantly reducing response time and enhancing overall system performance.

关键词： Scalability computational modeling Real-time systems computational efficiency Resource management Internet of Things Smart devices Artificial intelligence Optimization Edge computing

来源：评论

学校读者我要写书评

暂无评论

High-Performance File Searching with Rust: Parallelized Indexing for Enhanced computational efficiency

High-Performance File Searching with Rust: Parallelized Inde...

引用

Interdisciplinary Approaches in Technology and Management for Social Innovation (IATMSI), IEEE Conference on

作者： Snehil Sinha Prabir Kalwani Aayush Shah Jesleena Gonsalves Dept. of Computer Engineering Mukesh Patel School of Technology Management and Engineering SVKM’s NMIMS University Mumbai India

ISBN: (数字)9798331521691

ISBN: (纸本)9798331521707

Conventional file explorer tools have many limitations, from sluggish performance to cluttered user interfaces. The Windows File Explorer is often criticized for its slow performance, taking an exorbitant amount of time to load search results, especially for extensive file collections. In contrast, the Rust-based file searcher employs parallelization to rapidly parse the file system and build a comprehensive index. This index is stored in memory using a HashMap data structure, providing constant-time lookups. This study presents a high-performance alternative using the Rust programming language to address these shortcomings. The study shows the advantages of parallelization of search in Rust and then compares file search using Rust and traditional Windows file explorer on multiple parameters, including Search time, CPU usage, and Memory usage. The Rust-Based File Searcher provides faster results at around 80 to 180 milliseconds compared to 4-16 seconds using Windows File Search.

关键词： Technological innovation Computer languages File systems User interfaces Programming Data structures computational efficiency Indexing

来源：评论

学校读者我要写书评

暂无评论

computational efficiency analysis of Wu et al.'s fast modular multi-exponentiation algorithm

引用

APPLIED MATHEMATICS AND COMPUTATION 2007年第2期190卷 1848-1854页

作者： Sun, Da-Zhi Huai, Jin-Peng Sun, Ji-Zhou Zhang, Jia-Wan Tianjin Univ Sch Comp Sci & Technol Tianjin 300072 Peoples R China Beihang Univ Sch Comp Beijing 100083 Peoples R China

Very recently, for speeding up the computation of modular multi-exponentiation, Wu et al. presented a fast algorithm combining the complement recoding method and the minimal weight binary signed-digit representation technique. They claimed that the proposed algorithm reduced the number of modular multiplications from 1.503k to 1.306k on average, where the value k is the maximum bit-length of two exponents. However, in this paper, we show that their claim is unwarranted. We analyze the computational efficiency of Wit et al.'s algorithm by modeling it as a Markov chain. Our main result is that Wu et al.'s algorithm requires 1.471k modular multiplications on average. (c) 2007 Elsevier Inc. All rights reserved.

关键词： modular multi-exponentiation complement recoding minimal weight binary signed-digit (BSD) representation computational efficiency Markov chain

来源：评论

学校读者我要写书评

暂无评论

computational efficiency of Meshfree Methods with Local-Coordinates Algorithm

引用

INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING 2015年第3期16卷 547-556页

作者： Chan, Wei Xuan Son, HungSun Yoon, Yong-Jin Nanyang Technol Univ Dept Engn Mech Mech & Aerosp Engn Singapore 639798 Singapore ERI N Singapore 637141 Singapore Ulsan Natl Inst Sci & Technol Sch Mech & Nucl Engn Ulsan 689798 South Korea

In this study, meshfree methods with uniform nodal distribution and local-coordinates shape functions are investigated The proposed meshfree method can be used with various shape functions and is tested on a test patch with a Laplace governing equation and both essential and Neumann boundary conditions. It is shown to reduce the computational time dependency on the number of nodes by 36%. This reduction of dependency is significant as it reduces the computational time by several orders for analysis of problems requiring very fine nodal distribution. The meshfree method with uniform nodal distribution and local coordinates shape function reduces and converges to the perfectly uniform nodal distribution with finer nodal distributions. The technique is illustrated on a rectangular Kirchhoff-Love plate to show the practical use of the technique for allowing higher order shape function and finer nodal distribution to be used with multiple overlapping boundary conditions as well as on an electromagnetic problem to explain the uses of this technique on solving multiple similar problem cases with slight changes in geometry of boundary nodes. Overall, the present methodology provides a simple way to increase the computational efficiency of meshfree methods in range of an order while retaining many of its benefits.

关键词： Meshfree computational efficiency Local-coordinates Shape function

来源：评论

学校读者我要写书评

暂无评论

computational efficiency Maximization for UAV-Assisted MEC Networks With Energy Harvesting in Disaster Scenarios

引用

IEEE INTERNET OF THINGS JOURNAL 2024年第5期11卷 9004-9018页

作者： Khalid, Reda Shah, Zaiba Naeem, Muhammad Ali, Amjad Al-Fuqaha, Ala Ejaz, Waleed Lakehead Univ Dept Elect & Comp Engn Barrie Hub Barrie ON L4M 3X9 Canada COMSATS Univ Islamabad Dept Elect & Comp Engn Wah Campus Wah Cantt 47040 Pakistan Hamad Bin Khalifa Univ Qatar Fdn Div Informat & Comp Technol Coll Sci & Engn Doha Qatar

Recently, unmanned aerial vehicle (UAV)-assisted mobile-edge computing (MEC) networks are considered to provide effective and efficient solutions for disaster management. However, the limited size of end-user devices comes with the limitation of battery lives and computational capacities. Therefore, offloading, energy consumption, and computational efficiency are significant challenges for uninterrupted communication in UAV-assisted MEC networks. This article considers a UAV-assisted MEC network with energy harvesting (EH). To achieve this, we mathematically formulate a mixed-integer nonlinear programming problem to maximize the computational efficiency of UAV-assisted MEC networks with EH under disaster situations. A power-splitting architecture splits the source power for communication and EH. We jointly optimize user association, transmission power of user equipment (UE), task offloading time, and UAV's optimal location. To solve this optimization problem, we divide it into three stages. In the first stage, we adopt k-means clustering to determine the optimal locations of the UAVs. In the second stage, we determine user association. In the third stage, we determine the optimal power of UE and offloading time using the optimal UAV location from the first stage and the user association indicator from the second stage, followed by linearization and the use of the interior-point method to solve the resulting linear optimization problem. Simulation results for offloading, no-offloading, offloading-EH, and no-offloading-EH scenarios are presented with a varying number of UAVs and UEs. The results show the proposed EH solution's effectiveness in offloading scenarios compared to no-offloading scenarios in terms of computational efficiency, bits computed, and energy consumption.

关键词： Autonomous aerial vehicles Optimization Task analysis Internet of Things Energy consumption Resource management Energy harvesting computational efficiency disaster management energy harvesting (EH) interior-point linearization unmanned aerial vehicles (UAVs)

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：