检索结果-内蒙古大学图书馆

MP²SDA: Multi-Party parallelized Sparse Discriminant Learning

ACM TRANSACTIONS ON KNOWLEDGE DISCOVERY FROM DATA 2020年第3期14卷 26-26页

作者： Bian, Jiang Xiong, Haoyi Fu, Yanjie Huan, Jun Guo, Zhishan Univ Cent Florida 4328 Scorpius St Orlando FL 32816 USA Baidu Inc Big Data Lab 10 Shangdi 10th St Beijing 100094 Peoples R China Missouri Univ Sci & Technol 500W 15th St Rolla MO 65409 USA

Sparse Discriminant Analysis (SDA) has been widely used to improve the performance of classical Fisher's Linear Discriminant Analysis in supervised metric learning, feature selection, and classification. With the increasing needs of distributed data collection, storage, and processing, enabling the Sparse Discriminant Learning to embrace the multi-party distributed computing environments becomes an emerging research topic. This article proposes a novel multi-party SDA algorithm, which can learn SDA models effectively without sharing any raw data and basic statistics among machines. The proposed algorithm (1) leverages the direct estimation of SDA to derive a distributed loss function for the discriminant learning, (2) parameterizes the distributed loss function with local/global estimates through bootstrapping, and (3) approximates a global estimation of linear discriminant projection vector by optimizing the "distributed bootstrapping loss function" with gossip-based stochastic gradient descent. Experimental results on both synthetic and real-world benchmark datasets show that our algorithm can compete with the aggregated SDA with similar performance, and significantly outperforms the most recent distributed SDA in terms of accuracy and F1-score.

关键词： Sparse discriminant analysis multi-party distributed parallelized

来源：评论

学校读者我要写书评

暂无评论

Development of a parallelized 2D/2D-axisymmetric Navier-Stokes equation solver for all-speed gas flows

引用

COMPUTERS & FLUIDS 2011年第1期45卷 241-248页

作者： Hu, Meng-Hua Wu, Jong-Shinn Chen, Yen-Sen Natl Chiao Tung Univ Dept Mech Engn Hsinchu Taiwan Natl Space Org Natl Appl Res Labs Hsinchu Taiwan

A parallelized 2D/2D-axisymmetric pressure-based, extended SIMPLE finite-volume Navier-Stokes equation solver using Cartesians grids has been developed for simulating compressible, viscous, heat conductive and rarefied gas flows at all speeds with conjugate heat transfer. The discretized equations are solved by the parallel Krylov-Schwarz (KS) algorithm, in which the ILU and BiCGStab or GMRES scheme are used as the preconditioner and linear matrix equation solver, respectively. Developed code was validated by comparing previous published simulations wherever available for both low- and high-speed gas flows. Parallel performance for a typical 2D driven cavity problem is tested on the IBM-1350 at NCHC of Taiwan up to 32 processors. Future applications of this code are discussed briefly at the end. (C) 2010 Elsevier Ltd. All rights reserved.

关键词： parallelized Finite-volume Extended SIMPLE Navier-Stokes equation Conjugate heat transfer

来源：评论

学校读者我要写书评

暂无评论

Evaluating parallelized support vector regression and nearest neighbor regression with different input variations for estimating daily global solar radiation of the humid subtropical region in China

引用

INTERNATIONAL JOURNAL OF LOW-CARBON TECHNOLOGIES 2023年第18期18卷 95-110页

作者： Yu, Xiang Nanchang Inst Technol Prov Key Lab Water Informat Cooperat Sensing & Int 289 Tianxiang Rd Nanchang 330099 Jiangxi Peoples R China

Indirectly estimating global solar radiation is strongly nonlinear and needs to be addressed by machine learning. Sequentially developing a machine learning model can be very time consuming. Moreover, whether and how the exogenous meteorological, geographical and temporal variables affect regression accuracy still has not been well understood. This paper evaluates parallelized support vector regression (SVR) and nearest neighbor regression (NNR) models for estimating daily global solar radiation of the humid subtropical region in China using existing Python libraries on a multi-core central processing unit (CPU) and a graphical processing unit (GPU). Seven input variations are studied. Two variations are commonly adopted in literature, four variations contain meteorological, geographical and/or temporal features with bounded Pearson correlation coefficients (PCCs), and the other variation simply include all the available features. Experimental results demonstrate that: SVR and NNR are equally powerful for nonlinear regression, and the variation comprising features with absolute PCCs no less than 0.3 (i.e. just all the meteorological features) is able to achieve most accurate estimation;the GPU-parallelized SVR model can accelerate parameter calibration and prediction;compared with the CPU-parallelized and GPU-parallelized SVR models, the GPU-parallelized NNR model is much more efficient and rather more scalable with the increment of the number of data samples;and the CPU-parallelized NNR model consumes quite less parameter calibration time than the GPU-parallelized NNR model, owing to different methods adopted for determining distances and significant time wasted by the GPU-parallelized NNR model on repeatedly calculating required information during cross-validation.

关键词： global solar radiation support vector regression nearest neighbor regression parallelized

来源：评论

学校读者我要写书评

暂无评论

parallelized Forward Kinematics Using Product of Exponentials in PyTorch 31st

Parallelized Forward Kinematics Using Product of Exponential...

引用

31st International Conference on Robotics in Alpe-Adria-Danube Region (RAAD)

作者： Vrigazov, Hristo Yovchev, Kaloyan Sofia Univ St Kliment Ohridski Fac Math & Informat Sofia Bulgaria

ISBN: (纸本)9783031048722;9783031048708;9783031048692

In open-chain robots, the forward and inverse kinematics problems are foundational for manipulating the robot. The forward kinematics takes as input the position of each joint of the robot and computes the position and orientation of the robot's end-effector. Commonly used methods for solving the forward kinematics problem include the product of exponentials formula and the Denavit-Hartenberg (D-H) representation. In this paper, we focus on a specific Python implementation of the product of exponentials (POE) formula (modern robotics package), which is one of the most popular implementations used in production, with nearly 1000 stars on Github, and has been downloaded over 5000 times in the last year. This POE implementation does not natively support computation in parallel for multiple inputs. Also, the product of exponentials implementation cannot be used in machine learning models that use automatic differentiation, such as neural networks. This paper contributes a parallelized, automatically differentiable implementation of the modern robotics Python implementation using PyTorch, one of the most popular deep learning frameworks. We achieve this by introducing a natural extension using PyTorch tensors to the special orthogonal group so(3) and the special euclidean group se(3). The newly introduced extensions form a group themselves, and all operations needed for the POE formula can be implemented in them in parallel, leading to significant speedup on our hardware - both on CPU (over 200 times) and especially on GPU (over 1600 times), compared to a naive sequential implementation in Python.

关键词： Forward kinematics Product of exponentials Pytorch parallelized

来源：评论

学校读者我要写书评

暂无评论

QTBiCGSTAB Algorithm for Large Linear Systemand parallelized

引用

INTERNATIONAL JOURNAL OF GRID AND DISTRIBUTED COMPUTING 2014年第6期7卷 257-274页

作者： Xu, Xiaomei Zhang, Jilin Wan, Jian Zhou, Li Jiang, Ming Zhejiang ChineseMed Univ Dept Med Technol Hangzhou Zhejiang Peoples R China Hanzhou Dianzi Univ Dept Comp Sci Technol Hangzhou Zhejiang Peoples R China Zhejiang Prov Engn Ctr Media Data Cloud Proc & An Key Lab Complex Syst Modeling & Simulat Minist Educ Hangzhou Zhejiang Peoples R China

According to the traditional stabilized biconjugate gradient algorithm (BiCGSTAB) deficiency in data locality, this paper proposed a QTBiCGSTAB algorithm whose core idea is that recursively divides sparse matrix with quarter tree into sub-matrix and reorders them, to improve the hit ratio of cache and enhance the algorithm's efficiency. And the idea is good for algorithm being parallized, that is proved by the numerical experiments later. It mainly shows, firstly, QTBiCGSTAB algorithm is more efficiency than BiCGSTAB, and the speedup would reach 1: 330. The target division length would be influnced on the algorithm's performance;Secondly, for large linear system, parallized QTBiCGSTAB is more efficiency than serial's.

关键词： QTBiCGSTAB Data locality parallelized

来源：评论

学校读者我要写书评

暂无评论

The Research on parallelized Fast Trilateral Filter on GPU Acceleration

The Research on Parallelized Fast Trilateral Filter on GPU A...

引用

IEEE International Conference on Electronics, Communications and Control (ICECC)

作者： Li, Xujie Wenzhou Univ Intelligent Informat Syst Inst Wenzhou 325035 Peoples R China

ISBN: (纸本)9781457703218

This paper design and implement a parallel fast trilateral filter algorithm on GPU. The trilateral filter can be decomposed into two bilateral filters which are implemented by the parallelized bilateral filter on CUDA separately. The performance optimization strategies are intensively discussed. The occupancy of CUDA kernel for trilateral filter reaches 0.833 which is measured by Visual Profiler. The results indicate that more than two orders of magnitude speedup have been achieved in the GPU as compared to the CPU counterpart for trilateral filter (512 x 512 ***), which demonstrates the significance of parallelism.

关键词： trilateral filter parallelized CUDA GPU

来源：评论

学校读者我要写书评

暂无评论

Extracting Roads from Remote Sensing Images with the Aid of Path Morphology and parallelized Graph Cuts

Extracting Roads from Remote Sensing Images with the Aid of ...

引用

8th Symposium on Multispectral Image Processing and Pattern Recognition (MIPPR) - Automatic Target Recognition and Navigation

作者： Zhang, Zhulin Zhou, Shaoguang Hohai Uinvers Coll Earth Sci & Engn Sch Surveying & Mapping Sci & Engn Nanjing 210098 Jiangsu Peoples R China

ISBN: (纸本)9780819498038

While Graph Cuts are used for image segmentation, there exist two problems: how to get better initial information of foreground and background and how to improve the executing efficiency of Graph Cuts algorithm. To solve the first problem, path morphology and line segment matching algorithm are performed to get initial background information at the same time as getting initial foreground information, so non-road pixels similar with road pixels in gray value or texture are avoided being segmented as road points. To cope with the second problem, push-relabel strategy is chosen and its parallelized version based on NVIDIA CUDA platform is performed in this paper. Our strategy is performed on dense built-up area and suburban district and proved to be effective in both accuracy and efficiency.

关键词： Road Extraction Path Morphology Initial Information parallelized Graph Cuts

来源：评论

学校读者我要写书评

暂无评论

NASA Technical Reports Server (Ntrs) 20020073160: Support for Debugging Automatically parallelized Programs

引用

2017年

NASA Technical Reports Server (Ntrs) 20020073160: Support for Debugging Automatically parallelized Programs by NASA Technical Reports Server (Ntrs); published by

关键词： (ntrs) 20020073160: automatically computer programs debugging hood, robert jost, gabriele nasa technical reports server (ntrs) parallel computers parallel programming parallelized program verification (computers)

来源：评论

学校读者我要写书评

暂无评论

NASA Technical Reports Server (Ntrs) 20010094066: Support for Debugging Automatically parallelized Programs

引用

2017年

NASA Technical Reports Server (Ntrs) 20010094066: Support for Debugging Automatically parallelized Programs by NASA Technical Reports Server (Ntrs); published by

关键词： (ntrs) 20010094066: automatically computer program integrity computer programs computers debugging distributed memory hood, robert interprocessor communication jost, gabriele nasa technical reports server (ntrs) parallel processing (computers) parallelized program verification (computers) subroutines

来源：评论

学校读者我要写书评

暂无评论

NASA Technical Reports Server (Ntrs) 20010110187: Support for Debugging Automatically parallelized Programs

引用

2016年

NASA Technical Reports Server (Ntrs) 20010110187: Support for Debugging Automatically parallelized Programs by NASA Technical Reports Server (Ntrs); NASA Technical Reports Server (Ntrs); published by

关键词： (ntrs) 20010110187: automatically data transfer (computers) debugging fortran hood, robert jost, gabriele nasa technical reports server (ntrs) parallel processing (computers) parallel programming parallelized program verification (computers) software development tools

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：