• Journal of the Korean Institute of Intelligent Systems
• /
• v.5 no.2
• /
• pp.70-78
• /
• 1995

The identification of a fuzzy model using input-output data consists of two parts :Structure identification and parameter identification. In this paper an algorithm to identify those parameters and structures is suggested to solve the problems of the conventional methods. Given a set of input-output data, the consequent parameters are identified by the Hough transform and clustering method, each of which considers the linearity and continuity respectively. The gradient descent algorithm is used to fine-tune parameters of a fuzzy model. Finally, it is shown that this method is useful for the identification of a fuzzy model by simulation, where we only consider a single input and single output system.

• Journal of computational fluids engineering
• /
• v.3 no.1
• /
• pp.37-45
• /
• 1998

A new and efficient inverse design method based on the numerical optimization technique has been developed. The 2-D incompressible Navier-Stokes equations are solved for obtaining the objective functions and coupled with the optimization procedure to perform the inverse design. The steepest descent and the conjugate gradient method have been applied to find the searching direction. The golden section method was applied to compute the design variable intervals. It has been found that the airfoil and the pump impellers are well converged to their targeting shapes.

• Phonetics and Speech Sciences
• /
• v.7 no.4
• /
• pp.17-25
• /
• 2015

For blind source separation of convolutive mixtures, FDICA(Frequency Domain Independent Component Analysis) algorithms are generally used. Since FDICA algorithm such as Sawada FDICA, IVA(Independent Vector Analysis) works on the frequency bin basis with a natural gradient descent method, it takes much time to converge. In this paper, we propose a new method to improve convergence speed in FDICA algorithm. The proposed method reduces the number of iteration drastically in the process of natural gradient descent method by applying a weighted inner product constraint of unmixing matrix. Experimental results have shown that the proposed method achieved large improvement of convergence speed without degrading the separation performance of the baseline algorithms.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.25 no.3
• /
• pp.189-200
• /
• 2005

Ultrasonic inspection methods are widely used for detecting flaws in materials. One of the more popular methods involves the extraction of an appropriate set of features followed by the use of a neural network for the classification of the signals in the feature space. This paper describes an approach which uses LMS method to determine the coordinates of the ultrasonic probe followed by the use of SAFT to estimate the location of the ultrasonic reflector The method is employed for classifying NDE signals from the steam generator tubes in a nuclear power plant. The classification results using this scheme for the ultrasonic signals from cracks and deposits within steam generator tubes are presented.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2012.04a
• /
• pp.157-162
• /
• 2012

A structural modification of baseframe is an effective method to avoid resonance in marine diesel generator (D/G) set which consists of diesel engine, generator and baseframe. However the reinforcement with thick plates or additional parts to increase the natural frequency can be less effective because of increased weight. Especially fine control of target mode based on the experience is difficult because the weight and interference of system have to be considered. In this paper, the design optimization of baseframe was performed to reduce the resonant vibration using a gradient descent method. The design parameters such as thickness, shape and location of baseframe parts are optimized to increase the torsional natural frequency of D/G set. From the actual test, the new designed baseframe reduced the vibration level in resonance by 55% without any increase of weight and interference. interference.

• Journal of Korea Multimedia Society
• /
• v.21 no.4
• /
• pp.471-477
• /
• 2018

As an important key technology using on electronic devices, face recognition has become one of the hottest technology recently. The traditional BP Neural network has a strong ability of self-learning, adaptive and powerful non-linear mapping but it also has disadvantages such as slow convergence speed, easy to be traversed in the training process and easy to fall into local minimum points. So we come up with an algorithm based on BP neural network but also combined with the PCA algorithm and other methods such as the elastic gradient descent method which can improve the original network to try to improve the whole recognition efficiency and has the advantages of both PCA algorithm and BP neural network.

• Communications of the Korean Mathematical Society
• /
• v.9 no.2
• /
• pp.439-448
• /
• 1994

It is known that the steepest descent(SD) method and the conjugate gradient(CG) method [1, 2, 5, 6] converge when these methods are applied to solve linear systems of the form Ax = b, where A is symmetric and positive definite. For some finite difference discretizations of elliptic problems, one gets positive definite matrices that are almost symmetric. Practically, the SD method and the CG method work for these matrices. However, the convergence of these methods is not guaranteed theoretically. The SD method is also called Orthores(1) in iterative method papers. Elman [4] states that the convergence proof for Orthores($\kappa$), with $\kappa$ a positive integer, is not heard. In this paper, we prove that the SD method and the CG method converge when the $\iota$$^2$ matrix norm of the non-symmetric part of a positive definite matrix is less than some value related to the smallest and the largest eigenvalues of the symmetric part of the given matrix.(omitted)

• International Journal of Control, Automation, and Systems
• /
• v.3 no.1
• /
• pp.43-55
• /
• 2005

In this paper, a predictive control method using self-recurrent wavelet neural network (SRWNN) is proposed for chaotic systems. Since the SRWNN has a self-recurrent mother wavelet layer, it can well attract the complex nonlinear system though the SRWNN has less mother wavelet nodes than the wavelet neural network (WNN). Thus, the SRWNN is used as a model predictor for predicting the dynamic property of chaotic systems. The gradient descent method with the adaptive learning rates is applied to train the parameters of the SRWNN based predictor and controller. The adaptive learning rates are derived from the discrete Lyapunov stability theorem, which are used to guarantee the convergence of the predictive controller. Finally, the chaotic systems are provided to demonstrate the effectiveness of the proposed control strategy.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.7A
• /
• pp.978-985
• /
• 2000

Blind equalizers recover the transmitted data using statistical characteristics of the signal alone. Among many alternatives, steepest gradient descent type algorithms such as the CMA and Sato algorithm are most widely utilized in practice. In this paper we propose a dual-mode NCMA algorithm, which combines the advantages of the dual mode CMA and Normalized CMA (NCMA) with the dual mode phase recovery algorithm. In addition, we perform computer simulations to demonstrate the performance improvement of the proposed algorithm with a QAM system. Simulation results show that the presented algorithm has a faster convergence speed and smaller steady-state residual error than the CMA and dual-mode CMA.

• Journal of Electrical Engineering and information Science
• /
• v.2 no.6
• /
• pp.191-196
• /
• 1997

Fuzzy clustering has been playing an important role in solving many problems. Fuzzy c-Means(FCM) algorithm is most frequently used for fuzzy clustering. But some fixed point of FCM algorithm, know as Tucker's counter example, is not a reasonable solution. Moreover, FCM algorithm is impossible to perform the on-line learning since it is basically a batch learning scheme. This paper presents unsupervised learning networks as an attempt to improve shortcomings of the conventional clustering algorithm. This model integrates optimization function of FCM algorithm into unsupervised learning networks. The learning rule of the proposed scheme is a result of formal derivation based on the gradient descent procedure of a fuzzy objective function. Using the result of formal derivation, two algorithms of fuzzy cluster analysis, the batch learning version and on-line learning version, are devised. They are tested on several data sets and compared with FCM. The experimental results show that the proposed algorithms find out the reasonable solution on Tucker's counter example.