检索结果-内蒙古大学图书馆

Operations and learning in neural networks for robust prediction

IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART B-CYBERNETICS 1997年第3期27卷 402-411页

作者： Kogiantis, AG PapantoniKazakos, T UNIV ALABAMA DEPT ELECT ENGNTUSCALOOSAAL 35487

We consider stochastic neural networks, the objective of which is robust prediction for spatial control. We develop neural structures and operations, in which the representations of the environment are preprocessed and provided in quantized format to the prediction layer, and in which the response of each neuron is binary. We also identify the pertinent stochastic network parameters, and subsequently develop a supervised learning algorithm for them. The on-line learning algorithm is based on the Kullback-Leibler performance criterion, it induces backpropagation, and guarantees fast convergence to the prediction probabilities induced by the environment, with probability one.

关键词： Intelligent networks Neural networks Robustness Stochastic processes Supervised learning Robust control Neurons backpropagation algorithms Convergence Robot control

来源：评论

学校读者我要写书评

暂无评论

An adaptive fuzzy-neural-network controller for ultrasonic motor drive using the LLCC resonant technique

引用

IEEE TRANSACTIONS ON ULTRASONICS FERROELECTRICS AND FREQUENCY CONTROL 1999年第3期46卷 715-727页

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

In this study an adaptive fuzzy-neural network controller (AFNNC) is proposed to control a rotary traveling wave-type ultrasonic motor (USM) drive system. The USM is derived by a newly designed, high-frequency, two-phase voltage source inverter using two inductances and two capacitances (LLCC) resonant technique. Then, because the dynamic characteristics of the USM are complicated and the motor parameters are time varying, an AFNNC is proposed to control the rotor position of the USM. In the proposed controller, the USM drive system is identified by a fuzzy-neural-network identifier (FNNI) to provide the sensitivity information of the drive system to an adaptive controller. The backpropagation algorithm is used to train the FNNI on line. Moreover, to guarantee the convergence of identification and tracking errors, analytical methods based on a discrete-type Lyapunov function are proposed to determine the varied learning rates of the FNNI and the optimal learning rate of the adaptive controller. In addition, the effectiveness of the adaptive fuzzy-neural-network (AFNN) controlled USM drive system is demonstrated by some experimental results.

关键词： rotor position backpropagation algorithms Liapunov function learning rate Network Controller adaptive controller motor drives ultrasonic motors drive train voltage source inverter parametric motor Controllers dynamic characteristics

来源：评论

学校读者我要写书评

暂无评论

ANALYSIS OF GRADIENT DESCENT LEARNING algorithms FOR MULTILAYER FEEDFORWARD NEURAL NETWORKS

引用

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS 1991年第8期38卷 883-894页

作者： GUO, H GELFAND, SB PURDUE UNIV SCH ELECT ENGNW LAFAYETTEIN 47907 USA

We investigate certain dynamical properties of gradient-type learning algorithms as they apply to multilayer feedforward neural networks. These properties are more related to the multilayer structure of the net than to the particular threshold units at the nodes. The analysis explains the empirical observation that the weight sequence generated by backpropagation and related stochastic gradient algorithms exhibits a long term dependence on the initial choice of weights, and also a continued growth and/or drift even long after the outputs have converged. The analysis is carried out in two steps. First a simplified deterministic algorithm is derived using a describing function-type approach. Next, an analysis of the simplified algorithm is performed by considering an associated ordinary differential equation (ODE). Some numerical examples are given to illustrate the analysis. There has been almost no analysis of the dynamical behavior of backpropagation and related algorithms for the training of multilayer nets;this paper represents a first step in that direction.

关键词： Multi-layer neural network Algorithm design and analysis Neural networks Feedforward neural networks backpropagation algorithms Neurons Stochastic processes Nonhomogeneous media Performance analysis Equations

来源：评论

学校读者我要写书评

暂无评论

A PARALLEL GENETIC/NEURAL NETWORK LEARNING ALGORITHM FOR MIMD SHARED-MEMORY MACHINES

引用

IEEE TRANSACTIONS ON NEURAL NETWORKS 1994年第6期5卷 900-909页

作者： HUNG, SL ADELI, H OHIO STATE UNIV COLL ENGNCOLUMBUSOH 43210

A new algorithm is presented for training of multilayer feedforward neural networks by integrating a genetic algorithm with an adaptive conjugate gradient neural network learning algorithm. The parallel hybrid learning algorithm has been implemented in C on an MIMD shared memory machine (Cray Y-MP8/864 supercomputer). It has been applied to two different domains, engineering design and image recognition. The performance of the algorithm has been evaluated by applying it to three examples. The superior convergence property of the parallel hybrid neural network learning algorithm presented in this paper is demonstrated.

关键词： Neural networks Machine learning Biological neural networks backpropagation algorithms Genetic algorithms Image recognition Design engineering Convergence Feedforward neural networks Supercomputers

来源：评论

学校读者我要写书评

暂无评论

Comparative Analysis of the Results of Training a Neural Network with Calculated Weights and with Random Generation of the Weights

引用

AUTOMATION AND REMOTE CONTROL 2020年第7期81卷 1211-1229页

作者： Geidarov, P. Sh. Azerbaijan Natl Acad Sci Inst Control Syst Baku Azerbaijan

Neural networks based on metric recognition methods allow, based on the initial conditions of the computer vision task such as the number of images and samples, to determine the structure of the neural network (the number of neurons, layers, connections), and also allow to analytically calculate the values of the weights on the connections of the neural network. As feedforward neural networks, they can also be trained by classical learning algorithms. The possibility of precomputation of the values of the neural network weights allows us to say that the procedure for creating and training a feedforward neural network is accelerated in comparison with the classical scheme for creating and training a neural network where values of the weights are randomly generated. In this work, we conduct two experiments based on the handwritten numbers dataset MNIST that confirm this statement.

关键词： neural networks metric recognition methods nearest neighbor method backpropagation algorithms random weight initialization

来源：评论

学校读者我要写书评

暂无评论

Three term backpropagation algorithm for classification problem

Three term backpropagation algorithm for classification prob...

引用

作者： Shamsuddin, Siti Mariyam Darus, Maslina Saman, Fadhlina Izzah Soft Computing Research Group Faculty of Computer Science and Information System Universiti Teknologi Malaysia Universiti Kebangsaan Malaysia

Standard backpropagation Algorithm (BP) usually utilizes two term parameters;Learning Rate and Momentum Factor β. Despite the general success of this algorithm, there are several drawbacks such as existence of local minima, slow rates of convergence and modification of algorithm requires complex computations. In this study, further analysis of proportional factor (γ) for 3-Term BP is investigated on various scales of datasets;small, medium and large. Experiments are conducted using three UCI dataset;Balloon, Iris and Cancer. The results show that the 3-Term BP outperforms standard BP for small scale data, but does not work well for medium and large scale dataset. © ICS AS CR 2007.

关键词： backpropagation algorithms

来源：评论

学校读者我要写书评

暂无评论

A Theoretical Framework for End-to-End Learning of Deep Neural Networks With Applications to Robotics

引用

IEEE ACCESS 2023年 11卷 21992-22006页

作者： Li, Sitan Nguyen, Huu-Thiet Cheah, Chien Chern Nanyang Technol Univ Sch Elect & Elect Engn Singapore 639798 Singapore

Deep Learning (DL) systems are difficult to analyze and proving convergence of DL algorithms like backpropagation is an extremely challenging task as it is a highly non-convex and high-dimensional problem. When using DL algorithms in robotic systems, theoretical analysis of stability, convergence, and robustness is a vital procedure as robots need to operate in a predictable manner to ensure safety. This paper presents the first unified End-to-End (E2E) learning framework that can be applied to both classification problems and real-time kinematic robot control tasks. In the proposed forward simultaneous learning method, the weights of all layers of the fully connected neural networks are updated concurrently. The proposed E2E learning framework uses an adjustable ReLU activation function so that convergence of the output error to a bound, which is dependent on the neural network approximation error, can be ensured. In particular, it is shown that the error between ideal output and estimated output of the network can converge to and stay in a certain bound, and this bound reduces to zero when the approximation error is zero. Therefore, the robustness of the learning system can be ensured even in the presence of the approximation error. Two case studies are done on classification tasks using MNIST and CIFAR10 datasets by using the proposed learning method, and the results show that the E2E learning method achieves comparable test accuracy as the gradient descent method and the main advantage is that convergence can be ensured during training. The framework is also implemented on a UR5e robot with unknown kinematic model, which is the first result on real-time E2E learning of deep neural networks for kinematic control of robots with guaranteed convergence.

关键词： Robots Neural networks Real-time systems Deep learning backpropagation algorithms Kinematics Distance learning Electronic learning Deep neural networks end-to-end online learning

来源：评论

学校读者我要写书评

暂无评论

A NEURAL-NETWORK APPROACH TO PHOTOMETRIC STEREO INVERSION OF REAL-WORLD REFLECTANCE MAPS FOR EXTRACTING 3-D SHAPES OF OBJECTS

引用

IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS 1995年第9期25卷 1289-1300页

作者： RAJARAM, KV PARTHASARATHY, G FARUQI, MA UNIV QUEBEC ECOLE TECHNOL SUPERDEPT ELECT ENGNMONTREALPQ H3C 3P8CANADA

In this paper, we present a neural network approach to the problem of photometric stereo inversion of the reflectance maps of real-world objects for the purpose of estimating the 3-D attitudes of the surface patches of objects, As in the photometric stereo approach, here also the observation that there is a one-to-one mapping between the n-tuples of the photometric stereo image intensities and the orientations of the surface normals is valid, A multilayered feedforward neural network with backpropagation training algorithm is used as dimensionality reducer to effectively encode this mapping by associating the two components of surface normals to the observed intensities from three photometric stereo images of the underlying surface patches, The training patterns are sampled from the images of a Gaussian sphere of average reflectance containing both diffuse and specular components, The neural network thus trained has been tested on images of real-world objects with different shapes and reflectance properties, Using the surface normals estimated by the neural network, 3-D shapes of the objects have been reconstructed to a good approximation.

关键词： Neural networks Photometry Reflectivity Shape Multi-layer neural network Feedforward neural networks backpropagation algorithms Testing Surface reconstruction Image reconstruction

来源：评论

学校读者我要写书评

暂无评论

NEUROMORPHIC LEARNING OF CONTINUOUS-VALUED MAPPINGS FROM NOISE-CORRUPTED DATA

引用

IEEE TRANSACTIONS ON NEURAL NETWORKS 1991年第2期2卷 294-301页

作者： TROUDET, T MERRILL, W NASA LEWIS RES CTRCLEVELANDOH 44135

This paper analyzes the effect of noise on the learning performance of the backpropagation algorithm. A selective sampling of the training set is proposed to maximize the learning of continuous-valued mappings by backpropagation, when the data have been corrupted by noise. The training scheme is applied to the nonlinear control of a cart-pole system in the presence of noise. The neural computation provides the neurocontroller with good noise-filtering properties. In the presence of plant noise, the neurocontroller is found to be more stable than the teacher. A novel perspective on the application of neural network technology to control engineering is presented.

关键词： Neuromorphics backpropagation algorithms back propagation Neurocomputing control engineering Neurocontrollers Learning corrupted data Noise effects Networking Techniques nonlinear control

来源：评论

学校读者我要写书评

暂无评论

ON SEQUENTIAL CONSTRUCTION OF BINARY NEURAL NETWORKS

引用

IEEE TRANSACTIONS ON NEURAL NETWORKS 1995年第3期6卷 678-690页

作者： MUSELLI, M Consiglio Nazionale delle Ricerche Istituto per i Circuiti Elettronici Genoa Italy

A new technique, called sequential window learning (SWL), for the construction of two-layer perceptrons with binary inputs is presented. It generates the number of hidden neurons together with the correct values for the weights, starting from any binary training set. The introduction of a new type of neuron, having a window-shaped activation function, considerably increases the convergence speed and the compactness of resulting networks. Furthermore, a preprocessing technique, called hamming clustering (HC), is proposed for improving the generalization ability of constructive algorithms for binary feedforward neural networks. Its insertion in the sequential window learning is straightforward. Tests on classical benchmarks show the good performances of the proposed techniques, both in terms of network complexity and recognition accuracy.

关键词： Neural networks Neurons Feedforward neural networks Clustering algorithms backpropagation algorithms Network synthesis Convergence Testing Performance evaluation

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：