检索结果-内蒙古大学图书馆

Distributed quasi-newton algorithm for Non-Randomly Stored Data

JOURNAL OF SYSTEMS SCIENCE & COMPLEXITY 2025年 1-24页

作者： Liu, Xirui Wu, Mixia Liu, Bangshu Beijing Univ Technol Sch Math Stat & Mech Beijing 100124 Peoples R China Beijing Univ Technol Coll Econ & Management Beijing 100124 Peoples R China

Distributed learning is a well-established method for estimation tasks over extensively distributed datasets. However, non-randomly stored data can introduce bias into local parameter estimates, leading to significant performance degradation in classical distributed algorithms. In this paper, the authors propose a novel Distributed quasi-newton Pilot (DQNP) method for distributed learning with non-randomly distributed data. The proposed approach accommodates both randomly and non-randomly distributed data settings and imposes no constraints on the uniformity of local sample sizes. Additionally, it avoids the need to transfer the Hessian matrix or compute its inversion, thereby greatly reducing computational and communication complexity. The authors theoretically demonstrate that the resulting estimator achieves statistical efficiency under mild conditions. Extensive numerical experiments on synthetic and real-world data validate the theoretical findings and illustrate the effectiveness of the proposed method.

关键词： Communication-efficient computation efficiency distributed inference non-randomly distributed data quasi-newton algorithm

来源：评论

学校读者我要写书评

暂无评论

3-D Inversion Based on the Particle Swarm Optimization-quasi-newton Hybrid algorithm for SOTEM

引用

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2023年 61卷

作者： Xue, Guoqiang He, Yiming Chen, Weiying Wu, Xin Song, Wanting Chinese Acad Sci Key Lab Mineral Resources Inst Geol & Geophys Beijing 100029 Peoples R China Chinese Acad Sci Innovat Acad Earth Sci Beijing 100029 Peoples R China Univ Chinese Acad Sci Coll Earth & Planetary Sci Beijing 100049 Peoples R China

The short offset transient electromagnetic method (SOTEM) has become one of the main methods for coal hydrogeology investigations and mineral explorations due to its performance advantages of high precision, large depth abilities, and high efficiency. At the present time, due to the computational complexity and memory requirements for 3-D SOTEM inversion algorithms, SOTEM data inversion is mainly based on 1-D, which has made it difficult to meet fine detection requirements under complex geological conditions. In this study, a particle swarm optimization-quasi-newton hybrid algorithm (PSO-QN) was proposed, with high precision and fast 3-D inversions of SOTEM data achieved. In addition, a flight strategy was introduced in order to ensure the global optimization of individual particles, and a weighting strategy was adopted to avoid falling into local minimums. The 3-D inversions of the SOTEM data were tested based on a variety of geoelectric models. The results showed that the dependence of the PSO-QN hybrid algorithm on the number of particle swarms had dropped to two-thirds. Furthermore, the calculation time was significantly reduced to one-third. The inversion results of the field data revealed that the proposed inversion method possessed a strong anti-noise capacity, high speed, good stability, and high resolution.

关键词： Flight strategy particle swarm optimization (PSO) algorithm quasi-newton algorithm short offset transient electromagnetic method (SOTEM) weighting strategy

来源：评论

学校读者我要写书评

暂无评论

A lower bound based smoothed quasi-newton algorithm for group bridge penalized regression

引用

COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION 2017年第6期46卷 4694-4707页

作者： Cao, Yongxiu Huang, Jian Jiao, Yuling Liu, Yanyan Zhongnan Univ Econ & Law Sch Math & Stat Wuhan Hubei Peoples R China Wuhan Univ Sch Math & Stat Wuhan Hubei Peoples R China Shanghai Univ Finance & Econ Sch Stat & Management Shanghai Peoples R China Univ Iowa Dept Stat & Actuarial Sci Iowa City IA 52242 USA

In this paper, we propose a lower bound based smoothed quasi-newton algorithm for computing the solution paths of the group bridge estimator in linear regression models. Our method is based on the quasi-newton algorithm with a smoothed group bridge penalty in combination with a novel data-driven thresholding rule for the regression coefficients. This rule is derived based on a necessary KKT condition of the group bridge optimization problem. It is easy to implement and can be used to eliminate groups with zero coefficients. Thus, it reduces the dimension of the optimization problem. The proposed algorithm removes the restriction of groupwise orthogonal condition needed in coordinate descent and LARS algorithms for group variable selection. Numerical results show that the proposed algorithm outperforms the coordinate descent based algorithms in both efficiency and accuracy.

关键词： Lower bound rule Group bridge Penalized least squares quasi-newton algorithm

来源：评论

学校读者我要写书评

暂无评论

Node Localization of Wireless Sensor Networks Based on Hybrid Bat-quasi-newton algorithm

引用

INTERNATIONAL JOURNAL OF ONLINE ENGINEERING 2015年第6期11卷 38-42页

作者： Sun, Shunyuan Xu, Baoguo Jiangnan Univ Minist Educ Sch Internet Things Engn Wuxi 214122 Peoples R China Jiangnan Univ Minist Educ Key Lab Adv Proc Control Light Ind Wuxi 214122 Peoples R China

Concerning the problem that the least square method in the third stage of DV-Hop algorithm has low positioning accuracy, a localization algorithm was proposed which is the fusion of hybrid bat-quasi-newton algorithm and DV-Hop algorithm. First of all, the Bat algorithm (BA) was improved from two aspects: firstly, the random vector beta was adjusted adaptively according to bats fitness so that the pulse frequency had the adaptive ability. Secondly, bats were guided to move by the average position of all the best individuals before the current iteration so that the speed had variable performance;Then in the third stage of DV-Hop algorithm the improved bat algorithm was used to estimate node location and then quasi-newton algorithm was used to continue searching for the node location from the estimated location as the initial searching point. The simulation results show that, compared with the traditional DV-Hop algorithm and the improved algorithm of DV-Hop based on bat algorithm (BADV-Hop). Positioning precision of the proposed algorithm increases about 16.5% and 5. 18%, and the algorithm has better stability, it is suitable for high positioning precision and stability situation.

关键词： Bat algorithm (BA) DV-Hop algorithm quasi-newton algorithm Wireless Sensor Networks (WSN)

来源：评论

学校读者我要写书评

暂无评论

A Variable Step-Size Regularization-Based quasi-newton Adaptive Filter for System Identification

引用

IEEE ACCESS 2024年 12卷 10059-10068页

作者： Bekrani, Mehdi Zayyani, Hadi Qom Univ Technol Fac Elect & Comp Engn Qom Iran

The family of least mean square (LMS) based adaptive filtering algorithms suffers from convergence performance limitation due to the sensitivity of such algorithms to the eigenvalue spread of the input correlation matrix. The quasi-newton family of adaptive filtering algorithms addresses this limitation, but its performance is restricted by the estimation accuracy of the correlation matrix inverse, especially for highly correlated input signals. Furthermore, the convergence rate and the steady-state performance of both LMS and quasi-newton families are thoroughly depending on their step-sizes. In this paper, a variable step-size regularized quasi-newton adaptive algorithm is proposed in the context of system identification. In this algorithm, inspired by the matrix inversion lemma, a modified regularized matrix inverse with a time-varying regularization is computed such that during the convergence, the contribution of matrix inverse in the weight update is reduced, resulting in a more noise-robust algorithm. The paper further provides a convergence analysis of the proposed quasi-newton algorithm, wherein a variable step-size is proposed to achieve a high initial convergence rate and a low steady-state error in the context of system identification applications.

关键词： Adaptive filter quasi-newton algorithm regularization system identification variable step-size

来源：评论

学校读者我要写书评

暂无评论

General Framework for Multi-Sensor Multi-Target Localization: Sparse Solution and Iterative Data Association

引用

IEEE SENSORS JOURNAL 2025年第9期25卷 15822-15829页

作者： Seyedin, Seyed Mohammad B. Behnia, Fereidoon Sharif Univ Technol Dept Elect Engn Tehran *** Iran

The multi-sensor multi-target localization (MS-MTL) faces significant challenges, including data association ambiguities and computational complexity of processing the measurements from multiple sensors. In this article, we propose a general framework for MS-MTL that can be utilized in any arbitrary localization system. The framework consists of three stages: sparse representation and solution, data association, and multiple single-target localization. In the first stage, a sparse representation approach provides rough estimates of target locations, which does not achieve the Cram & eacute;r-Rao lower bound (CRLB), most probably. In the second stage, the data association is performed by leveraging the rough locations to resolve ambiguities among targets based on maximum likelihood (ML) criteria. Finally, the third stage employs a quasi-newton algorithm to refine individual target locations, using the initial estimates from the first stage as starting points for optimization. Furthermore, the association and localization stages can be repeated multiple times using the final locations as the next rough locations to refine results. We have evaluated the effectiveness and computational efficiency of the proposed framework using a multiple-input multiple-output (MIMO) system that relies on bistatic range measurements.

关键词： Data association likelihood maximization multi-sensor multi-target localization (MS-MTL) quasi-newton algorithm sparse representation Data association likelihood maximization multi-sensor multi-target localization (MS-MTL) quasi-newton algorithm sparse representation

来源：评论

学校读者我要写书评

暂无评论

Research on Invalid Detection Data Model of Mine Catalytic Sensors Based on Machine Learning

引用

IEEE SENSORS JOURNAL 2023年第3期23卷 1925-1932页

作者： Bowen, Wang China Coal Technol Engn Grp Chongqing Res Chongqing 400030 Peoples R China

In order to solve the problem of nonlinear failure data output by the catalytic combustible sensor (later referred as a sensor) when working in the mine environment, this article proposes a backpropagation (BP) neural network nonlinear data filtering model based on the Levenberg-Marquardt (L-M) algorithm. The experimental analysis shows that this model has obvious advantages in training speed and error performance compared with the BP neural network model established by a quasi-newton algorithm and adaptive linear regression (lr) momentum gradient descent algorithm. In terms of generalization ability, this model has better generalization ability than the radial basis function (RBF) feedforward neural network model with K -means clustering. Based on the above advantages, the model can effectively filter the sensor failure output data, eliminate the hidden danger of safety production caused by the failure output data, and improve the level of safety production in coal mines.

关键词： Gas detectors Sensor phenomena and characterization Biological neural networks Sensor systems Methane Resistance Data models Catalytic sensor backpropagation (BP) neural network combustible gas sensor Levenberg-Marquardt (L-M) algorithm machine learning neural network quasi-newton algorithm radial basis function (RBF) neural network

来源：评论

学校读者我要写书评

暂无评论

On the local convergence of a quasi-newton method for solving matrix polynomial equations

引用

APPLIED MATHEMATICS AND COMPUTATION 2023年 441卷

作者： Macias, E. M. Perez, R. Martinez, H. J. Univ Cauca Dept Math Popayan 190003 Colombia Univ Valle Dept Math Cali 76001 Colombia

In this article, we propose a quasi-newton algorithm to solve a matrix polynomial equa-tion, which can be seen as a generalization of the algorithm of the same type to solve the matrix quadratic equation proposed in Macias et al. (2016). The proposed algorithm re-duces the computational cost of the newton-Schur method traditionally used to solve this type of equations. We show that this algorithm is local and even quadratically convergent. Finally, we present numerical experiments that ratify the theoretical results developed.(c) 2022 Elsevier Inc. All rights reserved.

关键词： Matrix polynomial equations quasi-newton algorithm Local convergence

来源：评论

学校读者我要写书评

暂无评论

Acceleration of the EM algorithm

引用

WILEY INTERDISCIPLINARY REVIEWS-COMPUTATIONAL STATISTICS 2023年第6期15卷 e1618-e1618页

作者： Kuroda, Masahiro Geng, Zhi Okayama Univ Sci Dept Management 1-1 Ridaicho Okayama 7000005 Japan Beijing Technol & Business Univ Sch Math & Stat Beijing Peoples R China

The expectation-maximization (EM) algorithm is a well-known iterative algorithm for finding maximum likelihood estimates from incomplete data and is used in several statistical models with latent variables and missing data. The algorithm also exhibits a monotonic increase in a likelihood function and satisfies parameter constraints for its convergence. The popularity of the EM algorithm can be attributed to its stable convergence, simple implementation and flexibility in interpreting data incompleteness. Despite these computational advantages, the algorithm is linear convergent and suffers from very slow convergence when a statistical model has many parameters and a high proportion of missing data. Various algorithms have been proposed to accelerate the convergence of the EM algorithm. We introduce the acceleration of the EM algorithm using root-finding and vector extrapolation algorithms. The root-finding algorithms include Aitken's method and the newton-Raphson, quasi-newton and conjugate gradient algorithms. These algorithms with faster convergence rates allow the EM algorithm to be sped up. The vector extrapolation algorithms transform the sequence of estimates from the EM algorithm into a fast convergent sequence and can accelerate the convergence without modifying the EM algorithm. We describe the derivation of these acceleration algorithms and attempt to apply them to two examples. This article is categorized under: Statistical and Graphical Methods of Data Analysis > EM algorithm

关键词： acceleration of convergence conjugate gradient algorithm EM algorithm newton- Raphson algorithm quasi-newton algorithm squared extrapolation algorithm vector e algorithm vector extrapolation

来源：评论

学校读者我要写书评

暂无评论

Hydrocarbon production dynamics forecasting using machine learning: A state-of-the-art review

引用

FUEL 2023年第1期337卷

作者： Liang, Bin Liu, Jiang You, Junyu Jia, Jin Pan, Yi Jeong, Hoonyoung Southwest Petr Univ State Key Lab Oil & Gas Reservoir Geol & Exploitat Chengdu 610500 Peoples R China Chongqing Univ Sci & Technol Chongqing 401331 Peoples R China Res Inst Petr Explorat & Dev Beijing 100086 Peoples R China Seoul Natl Univ Res Inst Energy & Resources Inst Engn Res Dept Energy Syst Engn Seoul 08826 South Korea

Accurate prediction of hydrocarbon production is crucial for the oil and gas industry. However, the strong heterogeneity of underground formation, the inconsistency in oil-gas-water distribution, and the complex flow mechanisms make hydrocarbon production forecasting (HPF) difficult, which leads to a high level of uncertainty in the prediction results. The explosion of machine learning (ML) methodologies that are capable of analyzing big data shed new light on HPF using production data. In this article, an in-depth review is provided regarding HPF using ML methodologies. Firstly, the merits and drawbacks of traditional HPF methods are analyzed and summarized. Then, the applications of ML algorithms in HPF are reviewed in detail, especially concentrating on artificial neural network, support vector machine, and ensemble learning. For each algorithm, the basic theory and its variants are first introduced, and its applications in HPF are comprehensively demonstrated subsequently. Finally, this article presents the challenge and prospects of machine-learning-based HPF. Sophisticated ML proxy models can be con-structed and employed to deal with an extended type of input data such that improving the efficacy of data utilization. On the other hand, deep learning models designed to handle time-series data can gain more attention. Modeling approaches for multivariate time-series hydrocarbon production data using deep neural networks with similar functionality to LSTM may lead to more accurate and computationally efficient production forecasting.

关键词： AI Artificial intelligence ANN Artificial neural network SVM Support vector machine DCA Decline curve analysis MFNN Multilayer feed-forward neural networks MLP Multilayer perceptron BP Back Propagation algorithm DNN Deep neural network LSTM Long short-term memory GAN Generative adversarial network CNN Convolutional neural network FCNN Fully connected neural network RNN Recurrent neural network BPANN Back propagation artificial neural network NN Neural network cDC-GAN Conditional deep convolutional generative neural network CONV-1D 1D convolutional neural network GA Genetic algorithm DE Differential evolution SA Simulated annealing GSA Gravitational search algorithm PSO Particle swarm optimization MPA Marine predator algorithm ICA Imperialist competitive algorithm LM Levenberg Marquardt BRA Bayesian regularization algorithm MNN Multiple-neural-network HONN Higher-order neural network HOSO Higher-order synaptic operations MLMVN Multilayer neural network with multi-valued neurons DLSTM Deep long-short term memory EEMD Ensemble empirical mode decomposition FNN Feed-forward neural network BFGS quasi-newton algorithm ORF Oil recovery factor GOR Gas oil ratio ADAM Adaptive moment estimation method WOPR well oil production rate SGD stochastic gradient descent SVR Support vector regression LSSVM Least squares support vector machine RSM Response surface method LR Linear regression RF Random forest AdaBoost Adaptive boosting GBDT Gradient boosting decision tree GBM Gradient-boosting machine MORF Multi-objective random forest GBT Gradient boosting tree PCA Principal component analysis RREE Reverse recursive feature elimination

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：