检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Kanatsoulis, Charilaos I. Choi, Evelyn Jegelka, Stephanie Leskovec, Jure Ribeiro, Alejandro Dept. of Computer Science Stanford University United States School of Computation Information and Technology Technical University of Munich Computer Science & Artificial Intelligence Laboratory Massachusetts Institute of Technology United States Dept. of Electrical and Systems Engineering University of Pennsylvania United States

Positional encodings (PEs) are essential for effective graph representation learning because they provide position awareness in inherently position-agnostic transformer architectures and increase the expressive capacity of Graph Neural Networks (GNNs). However, designing powerful and efficient PEs for graphs poses significant challenges due to the absence of canonical node ordering and the scale of the graph. In this work, we identify four key properties that graph PEs should satisfy: stability, expressive power, scalability, and genericness. We find that existing eigenvector-based PE methods often fall short of jointly satisfying these criteria. To address this gap, we introduce PEARL, a novel framework of learnable PEs for graphs. Our primary insight is that message-passing GNNs function as nonlinear mappings of eigenvectors, enabling the design of GNN architectures for generating powerful and efficient PEs. A crucial challenge lies in initializing node attributes in a manner that is both expressive and permutation equivariant. We tackle this by initializing GNNs with random node inputs or standard basis vectors, thereby unlocking the expressive power of message-passing operations, while employing statistical pooling functions to maintain permutation equivariance. Our analysis demonstrates that PEARL approximates equivariant functions of eigenvectors with linear complexity, while rigorously establishing its stability and high expressive power. Experimental evaluations show that PEARL outperforms lightweight versions of eigenvector-based PEs and achieves comparable performance to full eigenvector-based PEs, but with one or two orders of magnitude lower complexity. Our code is available at https://***/ehejin/Pearl-PE. Copyright © 2025, The Authors. All rights reserved.

关键词： Encoding (symbols)

来源：评论

学校读者我要写书评

暂无评论

Robust 3D Face Alignment with Multi-Path Neural Architecture Search

Robust 3D Face Alignment with Multi-Path Neural Architecture...

引用

IEEE International Conference on Multimedia and Expo (ICME)

作者： Zhichao Jiang Hongsong Wang Xi Teng Baopu Li Institute of Deep Learning (IDL) Baidu Beijing China Department of Computer Science and Engineering Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications Southeast University Nanjing China Computer Vision Technology Institution Baidu Beijing China Baidu Research Baidu Sunnyvale USA

ISBN: (数字)9798350390155

ISBN: (纸本)9798350390162

3D face alignment is a very challenging and fundamental problem in computer vision. Existing deep learning-based methods manually design different networks to regress either parameters of a 3D face model or 3D positions of face vertices. However, designing such networks relies on expert knowledge, and these methods often struggle to produce consistent results across various face poses. To address this limitation, we employ Neural Architecture Search (NAS) to automatically discover the optimal architecture for 3D face alignment. We propose a novel Multi-path One-shot Neural Architecture Search (MONAS) framework that leverages multi-scale features and contextual information to enhance face alignment across various poses. The MONAS comprises two key algorithms: Multi-path Networks Unbiased Sampling Based Training and Simulated Annealing based Multi-path One-shot Search. Experimental results on three popular benchmarks demonstrate the superior performance of the MONAS for both sparse alignment and dense alignment.

关键词： Training Learning systems Solid modeling computer vision Three-dimensional displays Neural networks Simulated annealing Network architecture Search problems Faces

来源：评论

学校读者我要写书评

暂无评论

Service Continuity of Wind Energy Conversion System Based on Nonlinear Control and Fault Tolerant Technique

Service Continuity of Wind Energy Conversion System Based on...

引用

electrical, computer and Energy Technologies (ICECET), International Conference on

作者： Mokhtar Mahmoud Mohammedi Zoubir Gadouche Djemaa Nadjwa Boutaleb Azeddine Bendiabdellah Development (LDEE) of Electrical Department Laboratory of Electrical Drives University of Sciences & Technology of Oran Mohamed Boudiaf Oran Algeria Engineering (L2GEGI) of Electrical Department Laboratory of Electrical and Computer University Ibn Khaldun of Tiaret Tiaret Algeria

the work presented in this paper is about Wind Energy Conversion System (WECS) service continuity under Rotor Side Converter (RSC) IGBT open circuit fault during the application of Backstepping Control (BSC) to the Doubly Fed Induction Generator (DFIG) and during the application of the proposed fault tolerant technique to the RSC. Among the strong characteristics of the BSC technique is how to deal with DFIG-WECS nonlinearities in fault appearance degraded mode resulting in the extension of its functionality until the application of fault tolerant techniques. A new fault tolerant technique is associated to our DFIG-WECS in the aim of guaranteeing the continuity of service under IGBT open circuit fault presence. It is found that, the proposed voltage reconfiguration technique compared to the usual used fault tolerant techniques does not need to isolate the faulty leg resulting in a less inverter losses. The results obtained from the control technique and the reconfiguration voltages under MATLAB-Simulink gives value to the proposed techniques used through the faulty condition.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Multi-dimensional optical neural network

arXiv

引用

arXiv 2024年

作者： Jia, Zhetao Rubio, Hector Neim, Lilian Park, Jagang Preble, Stefan Kanté, Boubacar Department of Electrical Engineering and Computer Sciences University of California at Berkeley BerkeleyCA94720 United States Electrical and Computer Engineering Rochester Institute of Technology RochesterNY14623 United States Microsystems Engineering Rochester Institute of Technology RochesterNY14623 United States Materials Sciences Division Lawrence Berkeley National Laboratory BerkeleyCA94720 United States

The development of deep neural networks is witnessing fast growth in network size, which requires novel hardware computing platforms with large bandwidth and low energy consumption. Optical computing has been a potential candidate for next-generation computing systems. Specifically, wavelength-division multiplexing (WDM) has been widely adopted in optical neural network architecture to increase the computation bandwidth. Although existing WDM neural networks architectures have shown promise, they face challenges in the integration of light sources and further increase of the computing bandwidth. To overcome these issues, we introduce a mode-division multiplexing (MDM) strategy, offering a new degree of freedom in optical computing within the micro-ring resonator platform. We propose a MDM approach for small-scale networks and a multidimensional architecture for large-scale applications, supplementing WDM with MDM to enhance channel capacity for computations. In this work, we design and experimentally demonstrate key components for the MDM computing system, i.e., a multimode beam splitter, a thermo-optical tuner for the high-order mode, and a multimode waveguide bend. We further show a 2-by-2 matrix multiplexing system fabricated in a foundry that works for both MDM and MDM-WDM computing, which confirms that our approach successfully increases the input vector size for computing and ensures compatibility with existing WDM networks. Copyright © 2024, The Authors. All rights reserved.

关键词： Wavelength division multiplexing

来源：评论

学校读者我要写书评

暂无评论

Movable Frequency Diverse Array for Wireless Communication Security

引用

IEEE Transactions on Communications 2025年

作者： Cheng, Zihao Si, Jiangbo Li, Zan Liu, Pengpeng Huang, Yangchao Al-Dhahir, Naofal Xidian University State Key Laboratory of Integrated Services Networks Xian710071 China AFEU College of Information and Navigation China University of Texas at Dallas Department of Electrical and Computer Engineering RichardsonTX75080 United States

Frequency diverse array (FDA) is a promising antenna technology to achieve physical layer security by varying the frequency of each antenna at the transmitter. However, when the channels of the legitimate user and eavesdropper are highly correlated, FDA is limited by the frequency constraint and cannot provide satisfactory security performance. In this paper, we propose a novel movable FDA (MFDA) antenna technology where the positions of antennas can be dynamically adjusted in a given finite region. Specifically, we aim to maximize the secrecy capacity by jointly optimizing the antenna beamforming vector, antenna frequency vector and antenna position vector. To solve this non-convex optimization problem with coupled variables, we develop a two-stage alternating optimization (AO) algorithm based on block successive upper-bound minimization (BSUM) method. Moreover, to evaluate the security performance provided by MFDA, we introduce two benchmark schemes, i.e., phased array (PA) and FDA. Simulation results demonstrate that MFDA can significantly enhance security performance compared to PA and FDA. In particular, when the frequency constraint is strict, MFDA can further increase the secrecy capacity by adjusting the positions of antennas instead of the frequencies. © 1972-2012 IEEE.

关键词： Convex optimization

来源：评论

学校读者我要写书评

暂无评论

Correction: STASiamRPN: visual tracking based on spatiotemporal and attention

引用

Multimedia Systems 2022年第5期28卷 1557-1558页

作者： Wu, Ruixu Wen, Xianbin Liu, Zhanlu Yuan, Liming Xu, Haixia School of Computer Science and Engineering Tianjin University of Technology Tianjin China Key Laboratory of Computer Vision and System Ministry of Education Tianjin China Department of Physical Education Tianjin University of Technology Tianjin China

During online tracking, we use 5 frames as a video sequence, which includes the first 4 frames and the last frame. We use the last frame as the tracking frame. If the video sequence was less than 5 frames, we repeated...

关键词：

来源：评论

学校读者我要写书评

暂无评论

Augmenting Efficient Real-time Surgical Instrument Segmentation in Video with Point Tracking and Segment Anything

arXiv

引用

arXiv 2024年

作者： Wu, Zijian Schmidt, Adam Kazanzides, Peter Salcudean, Septimiu E. Robotics and Control Laboratory Department of Electrical and Computer Engineering The University of British Columbia VancouverBCV6T 1Z4 Canada Department of Computer Science Johns Hopkins University BaltimoreMD21218 United States

The Segment Anything Model (SAM) is a powerful vision foundation model that is revolutionizing the traditional paradigm of segmentation. Despite this, a reliance on prompting each frame and large computational cost limit its usage in robotically assisted surgery. Applications, such as augmented reality guidance, require little user intervention along with efficient inference to be usable clinically. In this study, we address these limitations by adopting lightweight SAM variants to meet the efficiency requirement and employing fine-tuning techniques to enhance their generalization in surgical scenes. Recent advancements in Tracking Any Point (TAP) have shown promising results in both accuracy and efficiency, particularly when points are occluded or leave the field of view. Inspired by this progress, we present a novel framework that combines an online point tracker with a lightweight SAM model that is fine-tuned for surgical instrument segmentation. Sparse points within the region of interest are tracked and used to prompt SAM throughout the video sequence, providing temporal consistency. The quantitative results surpass the state-of-the-art semi-supervised video object segmentation method XMem on the EndoVis 2015 dataset with 84.8 IoU and 91.0 Dice. Our method achieves promising performance that is comparable to XMem and transformer-based fully supervised segmentation methods on ex vivo UCL dVRK and in vivo CholecSeg8k datasets. In addition, the proposed method shows promising zero-shot generalization ability on the label-free STIR dataset. In terms of efficiency, we tested our method on a single GeForce RTX 4060/4090 GPU respectively, achieving an over 25/90 FPS inference speed. Code is available at: https://***/wuzijian1997/SIS-PT-SAM © 2024, CC BY.

关键词： Surgical equipment

来源：评论

学校读者我要写书评

暂无评论

Silicon Photonic Microresonator-Based High-Resolution Line-by-Line Pulse Shaping

arXiv

引用

arXiv 2024年

作者： Cohen, Lucas M. Wu, Kaiyi Myilswamy, Karthik V. Fatema, Saleha Lingaraju, Navin B. Weiner, Andrew M. School of Electrical and Computer Engineering Purdue University West LafayetteIN47907 United States The Johns Hopkins University Applied Physics Laboratory LaurelMD20723 United States

Optical pulse shaping stands as a formidable technique in ultrafast optics, radio-frequency photonics, and quantum communications. While existing systems rely on bulk optics or integrated platforms with planar waveguide sections for spatial dispersion, they face limitations in achieving finer (few- or sub-GHz) spectrum control. These methods either demand considerable space or suffer from pronounced phase errors and optical losses when assembled to achieve fine resolution. Addressing these challenges, we present a foundry-fabricated six-channel silicon photonic shaper using microresonator filter banks with inline phase control and high spectral resolution. Leveraging existing comb-based spectroscopic techniques, we devise a novel system to mitigate thermal crosstalk and enable the versatile use of our on-chip shaper. Our results demonstrate the shaper’s ability to phase-compensate six comb lines at tunable channel spacings of 3, 4, and 5 GHz. Specifically, at a 3 GHz channel spacing, we showcase the generation of arbitrary waveforms in the time domain. This scalable design and control scheme holds promise in meeting future demands for high-precision spectral shaping capabilities. Copyright © 2024, The Authors. All rights reserved.

关键词： Silicon photonics

来源：评论

学校读者我要写书评

暂无评论

Quantum noise of dark pulse microcombs

Quantum noise of dark pulse microcombs

引用

CLEO: QELS_Fundamental Science, QELS 2022

作者： Lao, Chenghao Jin, Xing Chang, Lin Xie, Weiqiang Shu, Haowen Wang, Xingjun Bowers, John E. Yang, Qi-Fan State Key Laboratory for Artificial Microstructure and Mesoscopic Physics Frontiers Science Center for Nano-Optoelectronics School of Physics Peking University Beijing100871 China Department of Electrical and Computer Engineering University of California Santa BarbaraCA93106 United States State Key Laboratory of Advanced Optical Communications System and Networks Department of Electronics School of Electronics Engineering and Computer Science Peking University Beijing100871 China

We investigate the fundamental timing jitter of dark pulse microcombs generated in an AlGaAs microresonator, where the quantum-limited frequency noise of the 91-GHz repetition rate is measured to be 0.5 Hz2/Hz. ©... 详细信息

ISBN: (纸本)9781557528209

关键词： Aluminum gallium arsenide

来源：评论

学校读者我要写书评

暂无评论

Recent Progress in Reduced Order Models for Electromagnetic Particle-in-Cell Kinetic Plasma Simulations

Recent Progress in Reduced Order Models for Electromagnetic ...

引用

Antennas and Propagation Society International Symposium

作者： I. Nayak J. L. Nicolini D.-Y. Na F. L. Teixeira Department of Electrical and Computer Engineering ElectroScience Laboratory The Ohio State University Columbus OH USA Department of Electrical Engineering Pohang University of Science and Technology (POSTECH) Pohang Gyeongbuk Province South Korea

We discuss recent progress on several model order reduction techniques for electromagnetic particle-in-cell (EM- PIC) simulations of kinetic plasmas: Proper Orthogonal De-composition (POD), Dynamic Mode Decomposition (DMD), and Koopman autoencoder (KAE). In the past, these techniques have been employed as powerful tools for data-driven modeling of nonlinear complex systems in general. For EMPIC kinetic plasma simulations, we show that they can reduce the number of degrees of freedom by projecting the solution space into a smaller space constituted of representative spatiotemporal modes. The trade- off between accuracy and cost can be controlled, in a tunable fashion, by adjusting the number of retained modes in the reduced model. POD, DMD, and KAE reduced order models are highly desirable to enable kinetic plasma simulations at a much reduced cost and hence more efficient parameter space sweeps during optimization loops for device design. These reduced order models are also useful as diagnosis tools for better understanding of the underlying plasma dynamics through the extraction of dominant features of the solution in a hierarchical fashion.

关键词：

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：