检索结果-内蒙古大学图书馆

HYDRA: BROKERING CLOUD AND HPC RESOURCES TO SUPPORT THE EXECUTION OF HETEROGENEOUS WORKLOADS AT SCALE

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Alsaadi, Aymen Jha, Shantenu Turilli, Matteo Department of Electrical and Computer Engineering Rutgers The State University of New Jersey PiscatawayNJ08854 United States Brookhaven National Laboratory UptonNY11973 United States

Scientific discovery increasingly depends on middleware that enables the execution of heterogeneous workflows on heterogeneous platforms. One of the main challenges is to design software components that integrate within the existing ecosystem to enable scale and performance across cloud and high-performance computing (HPC) platforms. Researchers are met with a varied computing landscape, which includes services available on commercial cloud platforms, data and network capabilities specifically designed for scientific discovery on government-sponsored cloud platforms, and scale and performance on HPC platforms. We present Hydra, an intra/cross cloud/HPC brokering system capable of concurrently acquiring resources from commercial/private cloud and HPC platforms and managing the execution of heterogeneous workflow applications on those resources. This paper offers four main contributions: (1) the design of brokering capabilities in the presence of task, platform, resource, and middleware heterogeneity;(2) a reference implementation of that design with Hydra;(3) an experimental characterization of Hydra’s overheads and strong/weak scaling with heterogeneous workloads and platforms;and (4) the implementation of a workflow that models sea-rise with Hydra and its scaling on cloud and HPC platforms. Copyright © 2024, The Authors. All rights reserved.

关键词： Middleware

Enhancing the Performance of Machine Learning Classification Models

学校读者我要写书评

暂无评论

Enhancing the Performance of Machine Learning Classification...

International Symposium on INFOTEH-JAHORINA (INFOTEH)

作者： Cihat Talat Akpinar Özcan Koşar Akif Durdu Computer Engineering Research and Development Elfatek Elektronik Ltd. Co. Konya Turkey Department of Electrical and Electronics Engineering Robotics Automation Control Laboratory (RAC-LAB) Konya Technical University Konya Turkey

ISBN: (数字)9798331515799

ISBN: (纸本)9798331515805

This study aims to compare and improve the performance of four different machine learning algorithms (Naive Bayes, Multi-Layer Perceptron (MLP), Decision Trees, and Support Vector Machines (SVM)) in classification problems. The analyses emphasize the impact of data preprocessing steps and hyperparameter optimization on model performance. As part of the data preprocessing, missing values were imputed, categorical data were transformed into numerical data, and normalization procedures were applied. It was observed that normalization significantly enhanced the performance of the MLP and SVM algorithms in particular. Furthermore, additional improvements in accuracy rates were achieved through hyperparameter optimization. Naive Bayes and Decision Trees were found to exhibit stable performance regardless of data scaling. This study demonstrates that proper data preprocessing and model selection can significantly enhance algorithm performance in classification problems.

关键词： Support vector machines Accuracy Machine learning algorithms Data preprocessing Hyperparameter optimization Data models Bayes methods Classification algorithms Numerical models Decision trees

Data-driven local operator finding for reduced-order modelling of plasma systems: II. Application to parametric dynamics

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Faraji, Farbod Reza, Maryam Knoll, Aaron Kutz, J. Nathan Plasma Propulsion Laboratory Department of Aeronautics Imperial College London London United Kingdom Department of Applied Mathematics and Electrical and Computer Engineering University of Washington Seattle United States

The dynamics of many real-world systems is dependent on certain parameters. These are either intrinsic to the system, inherently evolving over time, or that are external and enable controlling the system’s behavior. A fully generalizable model shall be able to reliably represent these parametric behaviors. Plasma technologies serve as an example of parametric dynamical systems. For instance, in a Hall thruster – an industrially important plasma propulsion technology for spacecrafts – the dominant plasma phenomena governing the device’s global dynamics change as the "self-sustained electric field" parameter varies over one characteristic cycle of the system. In addition, changing the control parameters, such as the intensity of externally applied magnetic field or the applied discharge voltage can majorly alter the dynamics of the thruster. In this Part II, we demonstrate that our novel data-driven local operator finding algorithm, Phi Method, which was discussed in Part I, can effectively learn the parametric dynamics so as to faithfully predict the systems’ behavior over unseen parameter spaces. We describe two adaptations of Phi Method toward parametric dynamics discovery, namely, the "parametric Phi Method" and the "ensemble Phi Method". Two demonstration cases are adopted: one, the 2D problem of a fluid flow past a cylinder, and the other, the 1D problem of Hall thruster plasma discharge. For the first test case, the "parametric Phi Method" is assessed comparatively against the parametric implementation of OPT-DMD. The predictive performance of the parametric Phi Method notably surpassed that of the "parametric OPT-DMD" in the fluid test case. Across both test cases, the parametric and ensemble Phi Method were rather equivalently well able to recover the governing parametric PDEs and provide accurate predictions over the test parameters. Analysis of the ensemble ROMs underlined that Phi Method learns the coefficients of the dominant terms in the dynamics with ve

关键词： Dynamic mode decomposition

Capacity Characterization for Binary RIS Aided MIMO Communication System Based on An Electromagnetic Scattering Model

学校读者我要写书评

暂无评论

Capacity Characterization for Binary RIS Aided MIMO Communic...

2023 International Applied Computational Electromagnetics Society Symposium, ACES-China 2023

作者： Bai, Xuyang Tan, Shurun Li, Ruihan Zhejiang University-University-University of Illinois Urbana-Champaign Institute Zhejiang Haining314400 China Zhejiang University State Key Laboratory of Modern Optical Instrumentation Hangzhou310027 China University of Illinois at Urbana-Champaign Department of Electrical and Computer Engineering UrbanaIL61801 United States

ISBN: (纸本)9781733509657

The reconfigurable intelligent surface (RIS) aided communication system has demonstrated its efficacy in enhancing information transmission efficiency without inducing excessive energy losses and costs. Analyzing the channel characteristics of reconfigurable intelligent surface (RIS) based communication systems using a rigorous electromagnetic model is crucial for the expedited development of modern communication modes such as high-frequency communication and near-field communication. Such analysis expectedly reveals potential communication degrees of freedom, which have yet to be widely explored. In this article, an electromagnetic model based approach is employed to estimate the received signal and channel matrix for a digital coded metasurface-aided MIMO communication system. The channel capacity of the system under random fixed RIS and controllable RIS is analyzed, respectively. The effect of random RIS is estimated by averaging the capacity under each fixed RIS. On the other hand, the channel capacity of the controllable RIS-aided system is obtained by traversing each RIS state and selecting the highest value. Our calculation results obtained at different RIS locations highlight that RIS improves the overall channel capacity of the original system, with particularly greater improvement observed in locations closer to the transceiver center. © 2023 Applied Computational Electromagnetics Society (ACES).

关键词： Channel capacity

THz wave regulates molecular conformational transition

学校读者我要写书评

暂无评论

THz wave regulates molecular conformational transition

IEEE International Conference on Plasma Science (ICOPS)

作者： S. Gu Z. Zhu Y. Li School of Optical-Electrical Computer Engineering University of Shanghai for Science and Technology Shanghai China Innovation Laboratory of Terahertz Biophysics National Innovation Institute of Defense Technology Beijing China

ISBN: (数字)9798350360912

ISBN: (纸本)9798350360929

Conversion of molecular conformations is closely related to many important biological processes. It follows that modulating the conformation of a molecular is a promising way to regulate its biological behavior. Recently terahertz (THz) waves have been proven as effective stimuli in altering the molecular dynamics as the transitions between vibrational or rotational energy levels of molecules follow the absorption of electromagnetic radiations in the far- or mid-infrared band (taken as the generalized THz band for convenience). There have been extensive efforts devoted into studying the rich bioeffects of THz waves and the underlying mechanisms [1–5]. Nevertheless, the electromagnetic frequencies are mostly in the relatively high-frequency THz band (approximately >30 THz). How the low-frequency THz waves affect the molecules remains elusive. In this work, we demonstrate through molecular dynamics simulations that specific low-frequency THz radiations can well switch the retinal in the rhodopsin among its dominant conformation and other secondary ones. The interconversion of the retinal conformations is uncovered due to resonance of the external THz fields with the retinal molecules. More specifically, the resonance absorption leads to an increase of the kinetic energy, promotes the chance of the retinal overcoming its free energy barriers, and ultimately makes the conformational transitions. In addition, it is noteworthy that for this work we propose to use the angular momenta based spectral analysis method to interpret the molecular rotations and successfully pinpoint the effective THz frequencies. The findings of this study expand the scope of THz technologies in manipulating molecular conformations, which could be widely applied in a variety of biological processes.

关键词： Absorption Biological processes Switches Retina Resonance Energy states Plasmas

YOLO: An Efficient Terahertz Band Integrated Sensing and Communications Scheme with Beam Squint

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Luo, Hongliang Gao, Feifei Lin, Hai Ma, Shaodan Vincent Poor, H. Department of Automation Tsinghua University Beijing China Department of Electrical and Electronic Systems Engineering Graduate School of Engineering Osaka Metropolitan University Osaka Sakai Japan State Key Laboratory of Internet of Things for Smart City The Department of Electrical and Computer Engineering University of Macau China Department of Electrical and Computer Engineering Princeton University PrincetonNJ United States

Using communications signals for dynamic target sensing is an important component of integrated sensing and communications (ISAC). In this paper, we propose to utilize the beam squint effect to realize fast non-cooperative dynamic target sensing in massive multiple input and multiple output (MIMO) Terahertz band communications systems. Specifically, we construct a wideband channel model of the echo signals, and design a beamforming strategy that controls the range of beam squint by adjusting the values of phase shifters and true time delay lines. With this design, beams at different subcarriers can be aligned along different directions in a planned way. Then the received echo signals at different subcarriers will carry target information in different directions, based on which the targets’ angles can be estimated through sophisticatedly designed algorithm. Moreover, we propose a supporting method based on extended array signal estimation, which utilizes the phase changes of different frequency subcarriers within different orthogonal frequency division multiplexing (OFDM) symbols to estimate the distances and velocities of dynamic targets. Interestingly, the proposed sensing scheme only needs to transmit and receive the signals once, which can be termed as You Only Listen Once (YOLO). Compared with the traditional ISAC methods that require time consuming beam sweeping, the proposed one greatly reduces the sensing overhead. Simulation results are provided to demonstrate the effectiveness of the proposed schemes. Copyright © 2023, The Authors. All rights reserved.

关键词： Orthogonal frequency division multiplexing

Radar Clutter Covariance Estimation: A Nonlinear Spectral Shrinkage Approach 48

学校读者我要写书评

暂无评论

Radar Clutter Covariance Estimation: A Nonlinear Spectral Sh...

48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023

作者： Jain, Shashwat Krishnamurthy, Vikram Rangaswamy, Muralidhar Kang, Bosung Gogineni, Sandeep Cornell University Electrical and Computer Engineering NY United States Wright Patterson Airforce Base Air Force Research Laboratory OH United States University of Dayton Research Institute DaytonOH United States Information Systems Laboratories Inc. DaytonOH United States

ISBN: (纸本)9781728163277

In this paper, we exploit the spiked covariance structure of the clutter plus noise covariance matrix for adaptive radar signal processing. Using state-of-the-art techniques from mathematical finance and high dimensional statistics we propose a non-linear shrinkage-based rotation invariant spiked covariance matrix estimator. We compare the proposed estimator with Rank Constrained Maximum Likelihood (RCML)-Expected Likelihood (EL) covariance estimator using the Challenge dataset generated from RFView. We demonstrate that the computation-time for the proposed estimator is less than the RCML-EL estimator with identical Signal to Clutter plus Noise (SCNR) performance for the Challenge dataset. We derive the lower bound and upper bound for the normalized SCNR and empirically show that RCML-EL and the proposed estimator perform within these derived bounds for the Challenge dataset. We state the convergence for the spiked eigenvalues of the estimator. © 2023 IEEE.

关键词： Shrinkage

Transfer Learning for Passive Sonar Classification using Pre-trained Audio and ImageNet Models

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Mohammadi, Amirmohammad Kelhe, Tejashri Carreiro, Davelle Van Dine, Alexandra Peeples, Joshua The Department of Electrical and Computer Engineering Texas A&M University College StationTX United States The Massachusetts Institute of Technology Lincoln Laboratory LexingtonMA United States

Transfer learning is commonly employed to leverage large, pre-trained models and perform fine-tuning for downstream tasks. The most prevalent pre-trained models are initially trained using ImageNet. However, their ability to generalize can vary across different data modalities. This study compares pretrained Audio Neural Networks (PANNs) and ImageNet pretrained models within the context of underwater acoustic target recognition (UATR). It was observed that the ImageNet pretrained models slightly out-perform pre-trained audio models in passive sonar classification. We also analyzed the impact of audio sampling rates for model pre-training and fine-tuning. This study contributes to transfer learning applications of UATR, illustrating the potential of pre-trained models to address limitations caused by scarce, labeled data in the UATR domain. © 2024, CC BY.

关键词： Sonar

Sample-Based Conservative Bias Linear Power Flow Approximations

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Buason, Paprapee Misra, Sidhant Molzahn, Daniel K. Los Alamos National Laboratory Los AlamosNM87545 United States School of Electrical and Computer Engineering Georgia Institute of Technology AtlantaGA30313 United States

The power flow equations are central to many problems in power system planning, analysis, and control. However, their inherent non-linearity and non-convexity present substantial challenges during problem-solving processes, especially for optimization problems. Accordingly, linear approximations are commonly employed to streamline computations, although this can often entail compromises in accuracy and feasibility. This paper proposes an approach termed Conservative Bias Linear Approximations (CBLA) for addressing these limitations. By minimizing approximation errors across a specified operating range while incorporating conservativeness (over- or under-estimating quantities of interest), CBLA strikes a balance between accuracy and tractability by maintaining linear constraints. By allowing users to design loss functions tailored to the specific approximated function, the bias approximation approach significantly enhances approximation accuracy. We illustrate the effectiveness of our proposed approach through several test cases. © 2024, CC BY-NC-ND.

关键词： Electric load flow