检索结果-内蒙古大学图书馆

Learning techniques to train neural networks as a state selector for inverter-fed induction machines using direct torque control

引用

IEEE TRANSACTIONS ON POWER ELECTRONICS 1997年第5期12卷 788-799页

作者： Cabrera, LA Elbuluk, ME Zinger, DS NO ILLINOIS UNIV DEPT ELECT ENGNDE KALBIL 60115

Neural networks are receiving attention as controllers for many industrial applications. Although these networks eliminate the need for mathematical models, they require a lot of training to understand the model of a plant or a process. Issues such as learning speed, stability, and weight convergence remain as areas of research and comparison of many training algorithms. This paper discusses the application of neural networks to control induction machines using direct torque control (DTC), A neural network is used to emulate the state selector of the DTC, The training algorithms used in this paper are the backpropagation, adaptive neuron model, extended Kalman filter, and the parallel recursive prediction error, Computer simulations of the motor and neural-network system using the four approaches are presented and compared, Discussions about the parallel recursive prediction error and the extended Kalman filter algorithms as the most promising training techniques is presented, giving their advantages and disadvantages.

关键词： adaptive neuron algorithm backpropagation algorithm direct torque control extended Kalman filter algorithm gradient descent algorithms induction machines neural-networks structure and training Newton algorithms parallel recursive prediction error algorithm

来源：评论

学校读者我要写书评

暂无评论

Automatic recognition of sleep spindles in EEG by using artificial neural networks

引用

EXPERT SYSTEMS WITH APPLICATIONS 2004年第3期27卷 451-458页

作者： Acir, N Güzelis, C Dokuz Eylul Univ Dept Elect & Elect Engn TR-35160 Izmir Turkey

In this paper, we introduce a two-stage procedure based on artificial neural networks for the automatic recognition of sleep spindles (SSs) in a multi-channel electroencephalographic signal. In the first stage, a discrete perceptron is used to eliminate definite non-SSs. The pre-classification done in the first stage not only reduces the computation time but also increases the overall detection performance of the procedure. In the second stage, the remaining SS candidates after pre-classification procedure are aimed to be separated from each other by an artificial neural network that would function as a post-classifier. Two different networks, i.e. a backpropagation multilayer perceptron and radial basis support vector machine (SVM), are proposed as the post-classifier and compared in terms of their classification performances. Visual evaluation, by two electroencephalographers (EEGers), of 19 channel EEG records of 6 subjects showed that the best performance is obtained with a radial basis SVM providing an average sensitivity of 94.6% and an average false detection rate of 4.0%. (C) 2004 Elsevier Ltd. All rights reserved.

关键词： EEG sleep spindle recognition support vector machines backpropagation algorithm pattern recognition

来源：评论

学校读者我要写书评

暂无评论

Design of an Under-Actuated Exoskeleton System for Walking Assist While Load Carrying

引用

ADVANCED ROBOTICS 2012年第5-6期26卷 561-580页

作者： Yu, S. N. Lee, H. D. Lee, S. H. Kim, W. S. Han, J. S. Han, C. S. Hanyang Univ Ansan 426791 Kyeonggi Do South Korea Korea Atom Energy Res Inst Taejon 305353 South Korea Hansung Univ Seoul 136792 South Korea

This study proposes an under-actuated wearable exoskeleton system to carry a heavy load. To synchronize that system with a user, a feasible modular-type wearable system and its corresponding sensor systems are proposed. The design process of the modular-type exoskeleton for lower extremities is presented based on the considered requirements. To operate the system with the user, human walking analysis and intention signal acquisition methods for actuating the proposed system are developed. In particular, a sensing data estimation strategy is applied to synchronize the exoskeleton system with a user correctly. Finally, several experiments were performed to evaluate the performance of the proposed exoskeleton system by measuring the electromyography signal of the wearer's muscles while walking on level ground and climbing up stairs with 20- to 40-kg loads, respectively. (C) Koninklijke Brill NV, Leiden and The Robotics Society of Japan, 2012

关键词： Wearable robot exoskeleton constant force mechanism muscle stiffness sensor backpropagation algorithm

来源：评论

学校读者我要写书评

暂无评论

Identification Of A Liquid Saturated Steam Heat Exchanger Using Focused Time Lagged Recurrent Neural Network Model

引用

IETE JOURNAL OF RESEARCH 2007年第1期53卷 69-82页

作者： Dudul, Sanjay V. St Gadge Baba Amravati Univ Fac Engn & Technol Dept Appl Elect Amravati 444602 India

[ABST].

关键词： Heat exchanger Multi-layer perceptron neural network backpropagation algorithm Focused time lagged recurrent neural network Gamma memory System identification

来源：评论

学校读者我要写书评

暂无评论

Analysis of two restart algorithms

引用

NEUROCOMPUTING 2006年第16-18期69卷 2301-2308页

作者： Lin, Wei Chen, Tianping Fudan Univ Inst Math Inst Nonlinear Math Sci Shanghai 200433 Peoples R China

Since the backpropagation algorithm used for neural network training suffers from a slow convergence and often sticking in local minima, the restart mechanism has been introduced, whose strategy is to cut off the training process and restart it with a fresh initialization when it seems unlikely to converge in a relatively short time. In this paper, we give detailed mathematical analysis on two versions of the restart algorithms. By deriving analytic expressions of the expected convergence time and the success rate, we illustrate why the restart algorithms work well and gain insights into the proper use of restarting. Numerical simulations are performed on the XOR problem, symmetry detection, parity problem and Arabic numeral recognition. We show the effectiveness of the restart algorithms, and compare them with simulated annealing. The analysis can also be applied to many other fields. (c) 2005 Elsevier B.V. All rights reserved.

关键词： restart mechanism neural network training backpropagation algorithm gradient descent method

来源：评论

学校读者我要写书评

暂无评论

Identification and control of rotary traveling-wave type ultrasonic motor using neural networks

引用

IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY 2001年第4期9卷 672-680页

作者： Lin, FJ Wai, RJ Hong, CM Chung Yuan Christian Univ Dept Elect Engn Chungli 32023 Taiwan

Neural networks (NNs) with varied learning rates are proposed to identify and control a nonlinear time-varying plant in this study. First, the network structure and the online learning algorithm of an NN are described, To guarantee the convergence of error states, analytical methods based on a discrete-type Lyapunov function are proposed to determine the varied learning rates of a three-layer NN with one hidden layer, Next, a rotary traveling-wave type ultrasonic motor (USM), which is driven by a newly designed high-frequency two-phase voltage source inverter using double indunctances double capacitances (LLCCs) resonant technique, is studied as an example of nonlinear time-varying plant to demonstrate the effectiveness of the proposed control system. Then, a robust control system is designed using two NNs to control the rotor position of the USM, In the proposed control system, the Jacobian of the USM drive system is identified by a neural-network identifier (NNI) to provide the sensitivity information to a neural-network controller (NNC), Moreover, the effectiveness of the NN controlled USM drive system is confirmed by some experimental results. Accurate tracking response can be obtained due to the powerful on-line learning capability of the NNs, Furthermore, the influence of uncertainties on the USM drive system can be reduced effectively.

关键词： backpropagation algorithm discrete-type Lyapunov function double indunctances double capacitances (LLCCs) resonant technique neural networks ultrasonic motor (USC) varied learning rates

来源：评论

学校读者我要写书评

暂无评论

POSSIBILITY AND NECESSITY PATTERN-CLASSIFICATION USING NEURAL NETWORKS

引用

FUZZY SETS AND SYSTEMS 1992年第3期48卷 331-340页

作者： ISHIBUCHI, H FUJIOKA, R TANAKA, H Department of Industrial Engineering College of Engineering University of Osaka Prefecture Osaka Japan

We propose two learning algorithms of neural networks for two-group discriminant problems from the view point of possibility and necessity. One algorithm corresponds to the possibility analysis and the other to the necessity analysis. The proposed algorithms are similar to the back-propagation algorithm and the difference stems from a formulation of a cost function to be minimized in each algorithm. Each cost function of the proposed algorithms is the weighted sum of squared errors, that is, the sum of squared errors with different penalties. When we discuss the possibility of Group 1, the penalty for the squared errors relating to the patterns in Group 1 is greater than that of Group 2. This means that, in the possibility analysis of Group 1, we attach greater importance to the patterns in Group 1 than to those in Group 2. On the other hand, when we discuss the necessity of Group 1, the penalty for the squared errors relating to the patterns in Group 2 is greater than that of Group 1.

关键词： NEURAL NETWORKS backpropagation algorithm PATTERN RECOGNITION APPROXIMATE CLASSIFICATION POSSIBILITY AND NECESSITY

来源：评论

学校读者我要写书评

暂无评论

PATTERN-RECOGNITION OF PARTIAL DISCHARGE IN XLPE CABLES USING A NEURAL NETWORK

引用

IEEE TRANSACTIONS ON ELECTRICAL INSULATION 1992年第3期27卷 543-549页

作者： SUZUKI, H ENDOH, T Power System Laboratory Hitachi Cable Limited Hitachi Japan

This paper describes an experimental study of pattern recognition of partial discharge (PD) in a crosslinked polyethylene (XLPE) cable by using a neural network (NN) system. The NN system was a three-layer artificial neural system with feedforward connections, and its learning method was a back-propagation algorithm incorporating an external teacher signal. Input information for the NN was a combination of the discharge magnitude, the number of pulse counts and the phase angle of applied voltage in which PD is produced. PD measurement was carried out using a PD pulse recorder for a 66 kV XLPE cable with an artificial defect under a 38 kV ac applied voltage. After learning 30 typical input patterns, the NN discriminated unknown patterns with 90% correct responses. The time duration including measuring time required for the NN to discriminate PD signal was approximately 30 s. In a long-term performance test of a 66 kV XLPE cable with an artificial defect, the NN-based alarm processor was able to recognize the presence of PD 1 h before breakdown of the cable, and successfully alerted the operator.

关键词： alarm systems cable insulation computerised pattern recognition high-voltage engineering insulation testing neural nets organic insulating materials partial discharges polymers power cables power engineering computing 66 kV XLPE cables alarm processor applied voltage artificial defect backpropagation algorithm discharge magnitude external teacher signal feedforward connections long-term performance test neural network number of pulse counts partial discharge phase angle Artificial neural networks backpropagation algorithms Cables Learning systems Neural networks Partial discharges Pattern recognition Polyethylene Pulse measurements Voltage

来源：评论

学校读者我要写书评

暂无评论

A SHUNTING MULTILAYER PERCEPTRON NETWORK FOR CONFUSING COMPOSITE PATTERN-RECOGNITION

引用

PATTERN RECOGNITION 1991年第11期24卷 1093-1103页

作者： WU, CH WANG, JF WU, WH NATL CHENG KUNG UNIV INST INFORMAT ENGNTAINANTAIWAN

In this paper, a shunting multilayer perceptron (SMLP) network trained with a modified backpropagation algorithm is proposed. In the SMLP, an MLP network shunts another MLP network in order to learn the relationship between any two training patterns. The SMLP network is endowed with two abilities in relation to patterns in complicated circumstances. Firstly, the SMLP network can perform confusing pattern recognition by weighting the node responses which are important for distinguishing two confusing patterns. Secondly, the SMLP network can perform composite pattern recognition by paying selective attention to the node responses which are distinctive for each individual pattern. The determination of the hidden node number for constructing a suitably sized SMLP network is also discussed here. For the experimental evaluations, ten spoken and handprinted digits and nine spoken and handprinted English alphabets were used. Comparisons to MLP, dynamic time warping algorithm, and nearest neighbor classifier showed satisfactory improvements in recognition accuracy for confusing/composite patterns.

关键词： NEURAL NETWORKS MULTILAYER PERCEPTRON backpropagation algorithm SPEECH PATTERN HANDPRINTED PATTERN CONFUSING COMPOSITE PATTERN

来源：评论

学校读者我要写书评

暂无评论

An Ensemble-Based Parallel Deep Learning Classifier With PSO-BP Optimization for Malware Detection

引用

IEEE ACCESS 2023年 11卷 76330-76346页

作者： Al-Andoli, Mohammed Nasser Sim, Kok Swee Tan, Shing Chiang Goh, Pey Yun Lim, Chee Peng Multimedia Univ Fac Engn & Technol Melaka 75450 Malaysia Multimedia Univ Fac Informat Sci & Technol Melaka 75450 Malaysia Deakin Univ Inst Intelligent Syst Res & Innovat Geelong Vic 3216 Australia

Digital networks and systems are susceptible to malicious software (malware) attacks. Deep learning (DL) models have recently emerged as effective methods to classify and detect malware. However, DL models often relies on gradient descent optimization in learning, i.e., the Back-Propagation (BP) algorithm;therefore, their training and optimization procedures suffer from several limitations, such as high computational cost and local suboptimal solutions. On the other hand, ensemble methods overcome the shortcomings of individual models by consolidating their strengths to increase performance. In this paper, we propose an ensemble-based parallel DL classifier for malware detection. In particular, a stacked ensemble learning method is developed, which leverages five DL base models and a neural network as a meta model. The DL models are trained and optimized with a hybrid optimization method based on BP and Particle Swarm Optimization (PSO) algorithms. To improve scalability and efficiency of the ensemble method, a parallel computing framework is exploited. The proposed ensemble method is evaluated using five malware datasets (namely, Drebin, NTAM, TOP-PE, DikeDataset, and ML_Android), and high accuracy rates of 99.2%, 99.3%, 98.7%, 100%, and 100% have been achieved, respectively. Its parallel implementation also significantly enhances the computational speed by a factor up to 6.75 times. These results ascertain that the proposed ensemble method is effective, efficient, and scalable, outperforming many other compared methods in malware detection.

关键词： Ensemble method malware detection deep learning parallel processing backpropagation algorithm particle swarm optimization

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：