• Journal of the Institute of Electronics and Information Engineers
• /
• v.52 no.8
• /
• pp.148-155
• /
• 2015

The Levenberg-Marquardt method is a well known solution about the least square problem. However, in a Target Motion Analysis(TMA) application most of researches have used the Gauss-Newton method as a batch estimator, which of inverse matrix calculation may causes instability problem. In this paper, Levenberg-Marquardt method is applied to TMA problem to prevent its divergence. In experiment, its performance is compared with Gauss-Newton in domain of range, course and speed. Monte Carlo simulation reveals the convergence time and reliability of the TMA based on Levenberg-Marquardt.

• Journal of applied mathematics & informatics
• /
• v.31 no.3_4
• /
• pp.373-378
• /
• 2013

In this paper, we consider two kinds of the Levenberg-Marquardt method for solve a system of nonlinear equations. We use line search on every iteration to guarantee that the Levenberg-Marquardt methods are globally convergent. Under mild conditions, we prove that while the de- scent condition can be satisfied at the iteration of the Levenberg-Marquardt method, the global convergence of the method can be established.

• Journal of applied mathematics & informatics
• /
• v.16 no.1_2
• /
• pp.183-194
• /
• 2004

In this paper, we consider a Levenberg-Marquardt method for the solution of constrained nonlinear equation problems. The global convergence is established even without requiring the existence of an accumulation point. Some numerical tests are also presented.

• Journal of the Korean GEO-environmental Society
• /
• v.10 no.5
• /
• pp.49-57
• /
• 2009

In this study, one of the complicated geotechnical problem, compression index was predicted by a artificial neural network method of Levenberg-Marquardt (LM) algorithm. Predicted values were compared and evaluated by the results of the Back Propagation (BP) method, which is used extensively in geotechnical engineering. Also two different results were compared with experimental values estimated by verified experimental methods in order to evaluate the accuracy of each method. The results from experimental method generally showed higher error than the results of both artificial neural network method. The predicted compression index by LM algorithm showed better comprehensive results than BP algorithm in terms of convergence, but accuracy was similar each other.

• Journal of the Korea Society of Computer and Information
• /
• v.14 no.8
• /
• pp.11-18
• /
• 2009

This paper proposes the way of improving learning speed in Levenberg-Marquardt algorithm using the principal submatrix of Jacobian matrix. The Levenberg-Marquardt learning uses Jacobian matrix for Hessian matrix to get the second derivative of an error function. To make the Jacobian matrix an invertible matrix. the Levenberg-Marquardt learning must increase or decrease ${\mu}$ and recalculate the inverse matrix of the Jacobian matrix due to these changes of ${\mu}$. Therefore, to have the proper ${\mu}$, we create the principal submatrix of Jacobian matrix and set the ${\mu}$ as the eigenvalues sum of the principal submatrix. which can make learning speed improve without calculating an additional inverse matrix. We also showed that our method was able to improve learning speed in both a generalized XOR problem and a handwritten digit recognition problem.

• Proceedings of the Earthquake Engineering Society of Korea Conference
• /
• 2002.03a
• /
• pp.20-27
• /
• 2002

Conventional Levenberg-Marquardt's nonlinear inversion method is simply modified by taking into account the second derivatives of the Hessian matrix so as to give robust inversion results. The weight of the second derivative terms is determined by the value of so-called λ in Levenberg-Marquardt's method. The new inversion method is applied to observed data from small-to-moderate earthquakes to simultaneously evaluate the modes parameters of the stochastic point-source model in and around the Korean Peninsula. Best estimates of the stochastic model parameters are obtained along with their statistics and compared with the previous results. Overall characteristics of the model parameters are found to be more of those of interplate than intraplate tectonic region.

• Journal of Korea Water Resources Association
• /
• v.46 no.4
• /
• pp.389-400
• /
• 2013

In this study, a conventional pipeline unsteady friction model has been integrated into Levenberg Marquardt method to calibrate friction coefficient in a pipeline system. The method of characteristics has been employed as the modeling platform for the frequency dependant model of unsteady friction. In order to obtain Hessian and Jacobian matrix for optimization, the direct differentiation of pressure to friction factor was calculated and sensitivities to friction for heads and discharges were formulated for implementation to the integration constant in the characteristic method. Using a hypothetical simple pipeline system, time series of pressure, introduced by a sudden valve closure, were obtained for various Reynolds numbers. Convergency in fiction factors were evaluated both in steady and unsteady friction models. The comparison of calibration performance between the proposed method and genetic algorithm indicates that faster and stabler behaviour of Levenberg Marquardt method than those of evolutionary calibration.

• Journal of the Korea Society of Computer and Information
• /
• v.15 no.4
• /
• pp.57-63
• /
• 2010

The damping parameter of Levenberg-Marquardt algorithm switches between error backpropagation and Gauss-Newton learning and affects learning speed. Fixing the damping parameter induces some oscillation of error and decreases learning speed. Therefore, we propose the way of a variable damping parameter with referring to the alternation of error. The proposed method makes the damping parameter increase if error rate is large and makes it decrease if error rate is small. This method so plays the role of momentum that it can improve learning speed. We tested both iris recognition and wine recognition for this paper. We found out that this method improved learning speed in 67% cases on iris recognition and in 78% cases on wine recognition. It was also showed that the oscillation of error by the proposed way was less than those of other algorithms.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.23 no.3
• /
• pp.364-373
• /
• 1999

A combination of conjugate gradient and Levenberg-Marquardt method is used to estimate the spatially varying scattering coefficient, ${\sigma}(r)$, in the solid and hollow spheres by utilizing the measured transmitted beams from the solution of an inverse analysis. The direct radiation problem associated with the inverse problem is solved by using the $S_{12}-approximation$ of the discrete ordinates method. The accuracy of the computations increased when the results from the conjugate gradient method are used as an initial guess for the Levenberg-Marquardt method of minimization. Optical thickness up to ${\tau}_0=3$ is used for the computations. Three different values of standard deviation are considered to examine the accuracy of the solution from the inverse analysis.

• Composites Research
• /
• v.31 no.5
• /
• pp.177-185
• /
• 2018

Fiber metal laminate (FML) is a type of hybrid composites which consist of metallic and fiber-reinforced plastic sheets. As the FML has a drawback of the delamination that is a failure of the interfacial adhesive layer, the nominal stresses and the energy release rates should be determined to identify the delamination behavior. However, it is difficult to derive the nominal stresses and the energy release rates since the operating temperature of the equipment is restricted. For this reason, the objective of this paper is to predict the mode-I nominal stress and the mode-I energy release rate of the adhesive layer using the inverse analysis based on the Levenberg-Marquardt method. First, the mode-I nominal stress was assumed as the tensile strength of the adhesive layer, and the mode-I energy release rate was obtained from the double cantilever beam test. Next, the finite element method was applied to predict the mode-I delamination behavior. Finally, the mode-I nominal stress and the mode-I energy release rate were predicted by the inverse analysis. In addition, the convergence of the parameters was validated by trying to input two cases of the initial parameters. Consequently, it is noted that the inverse analysis can predict the mode-I delamination behavior, and the two input parameters were converged to similar values.