检索结果-内蒙古大学图书馆

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Kaushik, Chiraag Romberg, Justin Muthukumar, Vidya Electrical and Computer Engineering Georgia Institute of Technology AtlantaGA30308 United States

The classical iteratively reweighted least-squares (IRLS) algorithm aims to recover an unknown signal from linear measurements by performing a sequence of weighted least squares problems, where the weights are recursively updated at each step. Varieties of this algorithm have been shown to achieve favorable empirical performance and theoretical guarantees for sparse recovery and p-norm minimization. Recently, some preliminary connections have also been made between IRLS and certain types of non-convex linear neural network architectures that are observed to exploit low-dimensional structure in high-dimensional linear models. In this work, we provide a unified asymptotic analysis for a family of algorithms that encompasses IRLS, the recently proposed lin-RFM algorithm (which was motivated by feature learning in neural networks), and the alternating minimization algorithm on linear diagonal neural networks. Our analysis operates in a "batched" setting with i.i.d. Gaussian covariates and shows that, with appropriately chosen reweighting policy, the algorithm can achieve favorable performance in only a handful of iterations. We also extend our results to the case of group-sparse recovery and show that leveraging this structure in the reweighting scheme provably improves test error compared to coordinate-wise reweighting. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

From Backscatter to Digital Twins: The Evolving Landscape of Additively Manufactured and Lens-Based RFIDs for Ultra-Long Range, sub-cm Accuracy, and Origami-Inspired Reconfigurability 14

From Backscatter to Digital Twins: The Evolving Landscape of...

引用

14th IEEE International Conference on RFID technology and Applications, RFID-TA 2024

作者： Al Jamal, Hani Joshi, Marvin Tentzeris, Manos M. School of Electrical and Computer Engineering Georgia Institute of Technology AtlantaGA United States

ISBN: (纸本)9798331540227

This paper reviews key advancements in RFID technology over the past decade and proposes future directions for its evolution. Notable progress has been made in additively manufactured and millimeter-wave (mm-wave) RFIDs, enhancing flexibility, scalability, and data throughput. Building on these innovations, we propose developments aimed at enabling ultra-long-range sensing, sub-centimeter localization, and high data rate transfer. We explore trends like origami-inspired reconfigurability, lens-based designs, and machine learning to improve RFID precision and adaptability. Additionally, we discuss the potential for fully energy-autonomous RFIDs and mm-wave technology for high-speed, long-range communication, positioning RFID as a central technology for next-generation IoT applications such as industrial IoT, virtual reality, smart cities, and autonomous vehicles. © 2024 IEEE.

关键词： Image segmentation

来源：评论

学校读者我要写书评

暂无评论

Solopulse Approach for SAR-with-DAR

Solopulse Approach for SAR-with-DAR

引用

2024 IEEE Conference on Computational Imaging Using Synthetic Apertures, CISA 2024

作者： Barnes, Christopher F. Truncale, Quentin X. Georgia Institute of Technology School of Electrical and Computer Engineering AtlantaGA United States

ISBN: (纸本)9798350308648

Solopulse radar signal processing can be characterized as a frequency-wavenumber or so-called 'omega-k' domain algorithm tailored specifically for digital array hardware with synthetic aperture software. Multiple data channels of a digital array radar permit image formation to commence with transmission of a single pulse, hence the moniker of 'solopulse.' Solopulse is applicable at short and long ranges, and with small and large scenes. With due consideration of flight geometry and pulse repetition interval designs, coherent integration of solopulse pixel domain data from a series of single-pulse images can result in progressively enhanced cross-range resolution. Pulse-to-pulse array movement need not be linear and may be randomized. Array step sizes that exceed Nyquist bounds result in grating lobes. Solopulse can also be characterized as wideband digital beamforming where angle-steered/range-focused receive-beam sets have a high spatial-packing density. Such 'regions-of-reconstruction' are obtained via spherical wave-field inversion through computationally efficient omega-k domain signal processing. Attributes of progressive resolution solopulse are explored and compared with range migration algorithm performance. © 2024 IEEE.

关键词： Radar imaging

来源：评论

学校读者我要写书评

暂无评论

Slow Convergence of Interacting Kalman Filters in Word-of-Mouth Social Learning 60

Slow Convergence of Interacting Kalman Filters in Word-of-Mo...

引用

60th Annual Allerton Conference on Communication, Control, and Computing, Allerton 2024

作者： Krishnamurthy, Vikram Rojas, Cristian R. School of Electrical and Computer Engineering Cornell University United States School of Electrical Engineering and Computer Science KTH Royal Institute of Technology Division of Decision and Control Systems Stockholm Sweden

ISBN: (纸本)9798331541033

We consider word-of-mouth social learning involving m Kalman filter agents that operate sequentially. The first Kalman filter receives the raw observations, while each subsequent Kalman filter receives a noisy measurement of the conditional mean of the previous Kalman filter. The prior is updated by the m-th Kalman filter. When m=2, and the observations are noisy measurements of a Gaussian random variable, the covariance goes to zero as k-1/3 for 7kappa;observations, instead of O(k-1) in the standard Kalman filter. In this paper we prove that for m agents, the covariance decreases to zero as k-(2m-1), i.e, the learning slows down exponentially with the number of agents. We also show that by artificially weighing the prior at each time, the learning rate can be made optimal as k-1. The implication is that in word-of-mouth social learning, artificially re-weighing the prior can yield the optimal learning rate. © 2024 IEEE.

关键词： Kalman filters

来源：评论

学校读者我要写书评

暂无评论

Investigating the Impacts of Device Geometry and an Alternative Write Current Scheme on Write Time and Switching Energy of SOT-MRAMs

Investigating the Impacts of Device Geometry and an Alternat...

引用

2024 IEEE Opportunity Research Scholars Symposium, ORSS 2024

作者： Kim, Seongjin Montello, William Narumanchi, Avanish Naeemi, Azad Haque Shazon, Md Nahid Georgia Institute of Technology School of Electrical and Computer Engineering AtlantaGA United States

ISBN: (纸本)9798350390698

The spin-orbit torque magnetoresistive random access memory (SOT-MRAM) is a next generation spintronic memory device known for its fast switching speed, high endurance and non-volatility. However, as this technology is scaled up and implemented in practical memory systems, issues in switching time, energy, and reliability must be addressed. This paper investigates the impact of device geometry on write performance and the implementation of a two-pulse spin current for magnetization switching as a way to achieve a faster and more energy-efficient write process for SOT-MRAMs. To this end, conduction current redistribution was first quantified using finite element simulation. Then, spin current was calculated using the spin drift-diffusion framework, and micromagnetic simulations were carried out for different device geometries to determine their write times and switching energies. It was revealed that up to 18% reduction in both switching energy and switching time can be achieved with a 5 nm β-W heavy metal layer thickness compared against a 3 nm one. Moreover, the two-pulse spin current magnetization switching scheme reduced the switching time by an average of 15%, while keeping the switching energy constant, compared to its single-pulse counterpart. © 2024 IEEE.

关键词： Magnetization

来源：评论

学校读者我要写书评

暂无评论

Q-learning based control for energy management of series-parallel hybrid vehicles with balanced fuel consumption and battery life

引用

Energy and AI 2023年第1期11卷 118-128页

作者： Saeid Ahmadian Mohammad Tahmasbi Reza Abedi Department of Mechanical Engineering Sharif University of TechnologyTehranIran Department of Electrical and Computer Engineering Concordia UniversityMontrealQCCanada Department of Mechanical Engineering K.N.Toosi University of TechnologyTehranIran

The present study investigates an energy management strategy based on reinforcement learning for seriesparallel hybrid vehicles. Hybrid electric vehicles allow using more advanced power management policies because of their complexity of power management. Towards this feature, a Q-Learning algorithm is proposed to design an energy management strategy. Compared to previous studies, an online reward function is defined to optimize fuel consumption and battery life cycle. Moreover, in the provided method, prior knowledge of the cycle and exact modeling of the vehicle are not required. The introduced strategy is simulated for four driving cycles in MATLAB software linked with ADVISOR. The simulation results show that in the HWFET cycle, the fuel consumption decreases by 1.25 %, and battery life increases by 65% compared to the rule-based method implemented in ADVISOR. Also, the results for the other driving cycles confirm the self-improvement property. In addition, it has been depicted that in the case of change in the driving cycle, the method performance has been maintained and gained better performance than the rule-based controller.

关键词： Hybrid electric vehicles Series-parallel configuration Power management policy Reinforcement learning Q-Learning

来源：评论

学校读者我要写书评

暂无评论

A Variable Gain Discrete Time Sliding Mode Observer for MMC Capacitor Voltage Estimation 7

A Variable Gain Discrete Time Sliding Mode Observer for MMC ...

引用

7th IEEE International electrical and Energy Conference, CIEEC 2024

作者： Lei, Yao Tianhao, Li Xinyu, Zhou Xuanang, Dong Yufei, Zhang College of Mechanical Engineering University of Shanghai for Science and Technology China School of Optical-Electrical and Computer Engineering University of Shanghai for Science and Technology China College of Information and Electrical Engineering China Agricultural University China

ISBN: (纸本)9798350359558

Modular Multilevel Converter (MMC) is one of the main development directions in the future medium and low voltage DC transmission and distribution field. In engineering, capacitive voltage sensors are usually configured for all submodules to ensure high reliability measurement, but this makes the hardware structure more complex. This paper proposes a control algorithm for online adjustment of observer gain based on discrete-time sliding mode observer (DTSMO). A weighted error judgment energy function is defined and three modes are distinguished. A transition mode is introduced between the approaching mode and sliding mode, and the gain within each mode is adjusted online according to the size of each component error function. This algorithm suppresses the chattering of traditional DTSMO during parameter changes and reduces the voltage error of each observed SM, this observer can eliminate all capacitive voltage sensors within a single bridge arm, and is designed based on the consideration of uncertain parameters outside of MMC. Its stability and convergence have been mathematically proven. Finally, simulation and experimental results of hardware prototypes has verified that the proposed VDTSMO observer has good steady-state and transient performance. © 2024 IEEE.

关键词： Bridges

来源：评论

学校读者我要写书评

暂无评论

Investigating Effective Connectivity within the Default Mode Network in Autism Spectrum Disorder: Insights from Dynamic Causal Modeling of Resting-State fMRI 31

Investigating Effective Connectivity within the Default Mode...

引用

31st National and 9th International Iranian Conference on Biomedical engineering, ICBME 2024

作者： Shafieizadegan, Soroor Shayegh, Farzaneh Varnosfaderani, Rassoul Amirfattahi Isfahan University of Technology Department of Electrical and Computer Engineering Isfahan84156-83111 Iran

ISBN: (纸本)9798331529710

Autism Spectrum Disorder (ASD) is characterized by deficiencies in social interactions and communications. The default mode network (DMN) is a resting-state network implicating in the social-cognitive deficits observed in ASD. Studying brain connectivity within the DMN in a large group can reveal useful insights into the mechanisms of ASD. This study investigated effective connectivity (EC) within the DMN, including the right and left middle temporal cortex (r/l MTC), right and left hippocampus, posterior cingulate cortex, medial prefrontal cortex, and right and left inferior parietal cortex (r/l IPC), using spectral dynamic causal modeling. A large sample size from the ABIDE database was used, comprising 1281 participants (577 ASD, aged 5-60 years;704 typically developing controls (TC), aged 5-57 years). Group-level analysis was conducted using Parametric Empirical Bayesian (PEB) method. The effects of sex, eye status during scanning, age, and Full-scale Intelligence Quotient (FIQ) on EC in ASD and TC groups were assessed. Results revealed significant group differences in EC from rIPC to IMTC, with decreased inhibition in ASD compared to TC. Additionally, a significant correlation was found between this EC and clinical scores (ADOS_G_TOTAL and ADOS_G_SOCIAL). Significant effects of sex and eye status on EC were observed in both groups, with differing patterns. In contrast, No association was found between age and EC, nor between FIQ and EC. In conclusion, this study highlights the alteration of EC in the DMN and underscores the importance of considering sex and eye status in EC studies of ASD. © 2024 IEEE.

关键词： Diseases

来源：评论

学校读者我要写书评

暂无评论

Precision in Motion: Comparative Analysis of YOLOvX Versions for Real-Time Multi-Object Tracking

Precision in Motion: Comparative Analysis of YOLOvX Versions...

引用

2024 International Conference on engineering and Computing, ICECT 2024

作者： Tariq, Javeria Fatima, Mehreen Kamal, Mohsin School of Electrical Engineering and Computer Sciences National University of Sciences and Technology Islamabad Pakistan

ISBN: (纸本)9798350349719

Real-Time object tracking is crucial in the world of computer vision. The real-world application of tracking in traffic management systems is quite challenging, and traditional methods have limitations due to speed and accuracy. This paper aims to explore You Only Look Once (YOLO)v8, a recent variant of YOLO, along with the deepsort tracking algorithm. Furthermore, a comparison is performed for YOLOv8 with key models-YOLOv3 and YOLOv4. Emphasizing accuracy, speed, and robustness, the study introduces new method at the crossroads of advanced object detection and dynamic tracking. We employed YOLOv8 on two datasets, COCO128 and a vehicle dataset, attaining Mean Average Precision (mAP) of 0.77 and 0.88 on the vehicle and COCO128 dataset, respectively. Rigorous evaluation provides valuable insights, guiding practical advancements in computer vision. © 2024 IEEE.

关键词： computer vision

来源：评论

学校读者我要写书评

暂无评论

Communication-Efficient Distributed Computing via Matrix Factorization 58

Communication-Efficient Distributed Computing via Matrix Fac...

引用

58th Asilomar Conference on Signals, Systems and computers, ACSSC 2024

作者： Hussein, Seif Razavikia, Saeed Daei, Sajad Fischione, Carlo School of Electrical Engineering and Computer Science Kth Royal Institute of Technology Stockholm Sweden

ISBN: (纸本)9798350354058

This paper presents a novel distributed computing framework, DCbMF (Distributed Computing by Matrix Factorization), for the efficient computation of linearly separable functions. Our framework operates within a multicast network of interconnected computing nodes to compute a set of output functions from input data through an efficient three-step process - Map, Shuffle, and Reduce. We cast the computation procedure as a sparse matrix factorization problem to achieve efficient communication and computation. To this end, we formulate an 0 optimization problem that seeks to minimize the number of nonzero elements in each matrix factor. Due to the intractability of 0 minimization, we turn to a relaxed 1 formulation of the problem. We devise a modified alternating direction method of multipliers to solve the biconvex optimization problem and prove the convergence of our algorithm to a stationary solution. The numerical experiments show that DCbMF outperforms similar computing frameworks for linearly separable function computation, achieving a substantial reduction in communication overhead by 98% while maintaining the same computation cost. Notably, leveraging sparse matrix factorization and alternating optimization highlights a fundamental tradeoff between compu-tation and communication costs and paves the way for scalable and efficient distributed computing applications. © 2024 IEEE.

关键词： Matrix factorization

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：