This paper compares the performance of Gradient descent with momentum & adaptive backpropagation (TRAINGDX) and BFGS quasi-Newton backpropagation (TRAINBFG) of backpropagationalgorithm in multilayer feed forward ...
详细信息
ISBN:
(纸本)9781467359863;9781467359870
This paper compares the performance of Gradient descent with momentum & adaptive backpropagation (TRAINGDX) and BFGS quasi-Newton backpropagation (TRAINBFG) of backpropagationalgorithm in multilayer feed forward Neural Network for Handwritten English Characters of Vowels. This analysis is done with five samples of Handwritten English Characters of Vowels collected from five different people and stored as an image. After partition these scanned image into 4 portions, the densities of these images are determined by using MATLAB function. An input pattern will use these 4 densities of each character as an input for the two different Neural Network architectures. In our proposed work the Multilayer feed forward neural networks will train with two learning algorithms;those are the variant of backpropagation learning algorithm namely Quasi-Newton backpropagation learning algorithm and Gradient descent with momentum and adaptive backpropagation learning algorithm for training set of the Handwritten English Characters of Vowels. The performance analysis of both Neural Network architectures is done for convergence and nonconvergence. Different observations have been considered for trends of error in the case of nonconvergence. From the observation of the result, it can be shown that in the performance of these above two learning algorithms with the training set of handwritten characters of Vowels, there is limitation of gradient descent learning algorithm for convergence due to the problem of local minima which is inherit problem of backpropagation learning algorithm.
This article presents the evaluation analysis of the radial function neural network embedded on an FPGA chip by experiments. The back-propagation algorithm has been embedded and tested for the feasibility for on-line ...
详细信息
ISBN:
(纸本)9781424421701
This article presents the evaluation analysis of the radial function neural network embedded on an FPGA chip by experiments. The back-propagation algorithm has been embedded and tested for the feasibility for on-line learning tasks. The nonlinear pattern classification task of the XOR logic has been conducted by the designed hardware. Performances are evaluated extensively by different orders of the Taylor-Maclaurin series expansion for approximating nonlinear functions and compared with results by the MATLAB program. The effects on the performance by the nonlinear function approximation have been analyzed by experimental studies of the XOR classification task.
backgroundSurgical simulation systems can be used to estimate soft tissue deformation during pre- and intra-operative planning. Such systems require a model that can accurately predict the deformation in real time. In...
详细信息
backgroundSurgical simulation systems can be used to estimate soft tissue deformation during pre- and intra-operative planning. Such systems require a model that can accurately predict the deformation in real time. In this study, we present a back-propagation neural network for predicting three-dimensional (3D) deformation of a phantom that incorporates the anatomy of the male pelvic region, i.e. the prostate and surrounding structures that support it. MethodIn the experiments and simulations, a needle guide is used to deform the rectal wall. The neural network predicts the deformation based on the relation between the undeformed and deformed shapes of the phantom. Training data are generated using a validated finite element (FE) model of the prostate and its surrounding structures. The FE model is developed from anatomically accurate magnetic resonance (MR) images. An ultrasound-based acoustic radiation force impulse imaging technique is used to measure in situ the shear wave velocity in soft tissue. The velocity is utilized to calculate the elasticities of the phantom. In the simulation study, the displacement and angle of the needle guide are varied. The neural network then predicts 3D phantom deformation for a given input displacement. ResultsThe results of the simulation study show that the maximum absolute linear and angular errors of the nodal displacement and orientation between neural network and FE predicted deformation are 0.03 mm and 0.01 degrees, respectively. ConclusionsThis study shows that a back-propagation neural network can be used to predict prostate deformation. Further, it is also demonstrated that a combination of ultrasound data, MR images and a neural network can be used as a framework for accurately predicting 3D prostate deformation in real time. Copyright (c) 2013 John Wiley & Sons, Ltd.
This article presents the hardware implementation of the Radial Basis Function (RBF) neural network whose internal weights are updated in the real-time fashion by the back-propagation algorithm. The floating-point pro...
详细信息
ISBN:
(纸本)9781424418206
This article presents the hardware implementation of the Radial Basis Function (RBF) neural network whose internal weights are updated in the real-time fashion by the back-propagation algorithm. The floating-point processor is designed on a field programmable gat array (FPGA) chip to execute nonlinear functions required in the parallel processing calculation of the back-propagation algorithm. The performance of the on-line learning process of the RBF chip is compared numerically with the results of the RBF neural network learning program written in the MATLAB software under the same condition to check the feasibility of the implemented neural chip. The performance of the designed RBF neural chip is tested for the real-time pattern classification of the nonlinear XOR logic.
This paper investigates a quantum neural network and discusses its application in control systems. A learning-type neural network-based controller that uses a multi-layer quantum neural network having qubit neurons as...
详细信息
This paper investigates a quantum neural network and discusses its application in control systems. A learning-type neural network-based controller that uses a multi-layer quantum neural network having qubit neurons as its information processing unit is proposed. Three learning algorithms;a back-propagation algorithm, a conjugate gradient algorithm and a real-coded genetic algorithm, are investigated to supervise the training of the multi-layer quantum neural network. To evaluate the learning performance and the capability of the quantum neural network-based controller, we conducted computational experiments for controlling a nonlinear discrete-time plant and a nonholonomic system - in this study a two-wheeled robot. The results of computational experiments confirm both the feasibility and the effectiveness of the quantum neural network-based controller and that the real-coded genetic algorithm is suitable for the learning method of the quantum neural network-based controller.
Cognitive Radios (CRs) are devices, which should be cognizant about the Spectrum Holes upon which the idea of a CR resides, which relates to the sensing and channel management function of the device. CR, therefore, mu...
详细信息
Cognitive Radios (CRs) are devices, which should be cognizant about the Spectrum Holes upon which the idea of a CR resides, which relates to the sensing and channel management function of the device. CR, therefore, must employ channel prediction techniques so as to decide the usage of channel and also prevents interference with the primary users (or incumbent users). In this paper, we use HMM to predict the channel usage patterns and to determine the channel occupancy states. We make use of BWA to train the parameters of the HMM model, Viterbi algorithm to estimate the hidden state of the channel and BWFA to predict the next state of the sequence. In addition to this, we compare the performance of the HMM-based model with that of a neural network-based predictor, which employs a time-delay feed forward network and which uses backpropagationalgorithm for training.
In this study, an artificial neural network (ANN) model using hybrid neural network is proposed for the design of aperture-coupled microstrip antennas (ACMSAs). The new hybrid model is developed by combining radial ba...
详细信息
In this study, an artificial neural network (ANN) model using hybrid neural network is proposed for the design of aperture-coupled microstrip antennas (ACMSAs). The new hybrid model is developed by combining radial basis function (RBF) and back-propagation algorithm (BPA). The performances evaluation of the hybrid model reveals superiority over the conventional BPA and RBF models in terms of error and time. The results obtained by the proposed model are compared with the simulation results obtained from the IE3D software package and also with the experimental results obtained from the fabricated ACMSA. The results show good agreement.
One-dimensional self-organizing maps neural network (SOM) is used in pile samples selection, and the outcome can improve the accuracy of back-propagation neural network (BP) is proved. Firstly, 71 pile samples are div...
详细信息
ISBN:
(纸本)9783037854846
One-dimensional self-organizing maps neural network (SOM) is used in pile samples selection, and the outcome can improve the accuracy of back-propagation neural network (BP) is proved. Firstly, 71 pile samples are divided into 5 groups according to SOM node weights. Each group is divided into training set, testing set, validation set to build 5 independent BP networks, called BP2. Secondly, 5 groups training set are merged into a new training set, similarly, a new test set and validation set to build another BP network, called BP1. At last, comparison of the performance of BP1 and BP2 show that using SOM network to select pile samples can build a BP network with better performance.
Fault diagnosis is an important problem in the process of chemical industry and the artificial neural network is widely applied in fault diagnosis of chemical process. A hybrid algorithm combining ant colony optimizat...
详细信息
ISBN:
(纸本)9783037855027
Fault diagnosis is an important problem in the process of chemical industry and the artificial neural network is widely applied in fault diagnosis of chemical process. A hybrid algorithm combining ant colony optimization (ACO) algorithm with back-propagation (BP) algorithm, also referred to as ACO-BP algorithm, is proposed to train the neural network weights and thresholds. The basic theory and steps of ACO-BP algorithm are given, and applied in fault diagnosis of the continuous stirred-tank reactor (CSTR). Experimental results prove that ACO-BP algorithm has good fault diagnosis precision, and it can detect the fault in CSTR promptly and effectively.
In this study, an inverse method is proposed for estimating the boundary conditions in a heat conduction problem using a regression analysis, neural network trained by a local optimizer and lastly, that trained by the...
详细信息
In this study, an inverse method is proposed for estimating the boundary conditions in a heat conduction problem using a regression analysis, neural network trained by a local optimizer and lastly, that trained by the local and global optimizers simultaneously. The test problem consists of a square slab with an internal heat source of circular shape. Once the boundary conditions of the square slab and periphery of the heat source are specified, the temperature can be estimated for two-dimensional heat conduction problems. This constitutes the forward heat transfer problem. Reverse heat transfer problem is formulated to determine the location of the heat source from some known temperature values elsewhere in the slab. A reasonably good solution is obtained from the said inverse problem using the proposed approach, whose performance is also compared with that of other two approaches.
暂无评论