• 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.

• Journal of applied mathematics & informatics
• /
• v.20 no.1_2
• /
• pp.35-59
• /
• 2006

In this paper, we generalized the results of [23, 26], and get the results of the condition number of the W-weighted Drazin-inverse solution of linear system W AW\chi=b, where A is an $m{\times}n$ rank-deficient matrix and the index of A W is $k_1$, the index of W A is $k_2$, b is a real vector of size n in the range of $(WA)^{k_2}$, $\chi$ is a real vector of size m in the range of $(AW)^{k_1}$. Let $\alpha$ and $\beta$ be two positive real numbers, when we consider the weighted Frobenius norm $\|[{\alpha}W\;AW,\;{\beta}b]\|$(equation omitted) on the data we get the formula of condition number of the W-weighted Drazin-inverse solution of linear system. For the normwise condition number, the sensitivity of the relative condition number itself is studied, and the componentwise perturbation is also investigated.

• Structural Engineering and Mechanics
• /
• v.12 no.5
• /
• pp.527-540
• /
• 2001

A new sort of learning algorithm named whole learning algorithm is proposed to simulate the nonlinear and dynamic behavior of RC members for the estimation of structural integrity. A mathematical technique to solve the multi-objective optimization problem is applied for the learning of the feedforward neural network, which is formulated so as to minimize the Euclidean norm of the error vector defined as the difference between the outputs and the target values for all the learning data sets. The change of the outputs is approximated in the first-order with respect to the amount of weight modification of the network. The governing equation for weight modification to make the error vector null is constituted with the consideration of the approximated outputs for all the learning data sets. The solution is neatly determined by means of the Moore-Penrose generalized inverse after summarization of the governing equation into the linear simultaneous equations with a rectangular matrix of coefficients. The learning efficiency of the proposed algorithm from the viewpoint of computational cost is verified in three types of problems to learn the truth table for exclusive or, the stress-strain relationship described by the Ramberg-Osgood model and the nonlinear and dynamic behavior of RC members observed under an earthquake.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 1997.10a
• /
• pp.262-269
• /
• 1997

The purpose of this study is to survey the shape finding of the plane truss structures under the prescribed displacement mode by the shape analysis. The shape analysis is peformed by the existence condition of a solution and Moore-Penrose generalized inverse matrix, and the prescribed displacement mode is the homologous deformation of structures. The shape analysis of structures is a kind of inverse problem and become the problem of a nonlinear equation. Newton-Raphson method is used to improve the accuracy of approximated solution. To prove the accuracy and the effectiveness of this method, four different shape examples are analyzed.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.27 no.11
• /
• pp.32-41
• /
• 1990

This paper proposes a method of using image features in serving a robot manipulator. Specifically, the con-cept 'feature' is first mathematically defined and then differential relationship between the robot motion and feature vector is derived in terms of Feature Jacobian Matrix and its generalized inverse. Also, by using more features than the number of DOFs of the robot, the visual servoing performance is shown to be improv-ed. Via various examples, the method of feature-based servoing of a robot proposed in this paper is proved to be effective for conducting object-oriented robotic tasks.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.42 no.8 s.338
• /
• pp.1-10
• /
• 2005

In this paper, the explicit low density codes construction from the generalized permutation matrices related to algebra theory is investigated, and we design several Jacket inverse block matrices on the recursive formula and permutation matrices. The results show that the proposed scheme is a simple and fast way to obtain the low density codes, and we also Proved that the structured low density parity check (LDPC) codes, such as the $\pi-rotation$ LDPC codes are the low density Jacket inverse block matrices too.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.20 no.3
• /
• pp.872-881
• /
• 1996

The concept of homology design has been devised for the application to large telescope structure by S.v.Hoerner. It is defined that the deformation of a structure shall be called homologous, if a given geometrical relation holds, for a given number of structural points, before, during, and after the deformation. Recently, the need of homology design in the structural design has been increase due to the required precision in the structure. Some researchers have utilized the theory on the structural design with finite element method in the late 1980s In the present investigation, a simple method using geometrical equality constraints is suggested to gain homologous deformation. The previous method is improved in that the decomposition of FEM eqation, which is very expensive, is not necessary. The basic formulations of the homology design with the optimization concept are described and several practical examples are solved to verify the usefulness and validity. Especially, a back-up structure of a satellite antenna is designed by the suggested method. The results are compared with those of existing researches.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 2001.04a
• /
• pp.251-258
• /
• 2001

Cable and membrane structures can be classified as a unstable structure in the view point of shape determination process. An unstable stucture at the initial state generally cannot take a role as the resistance for the external force. Therefore, there should be a stabilizing process to get the stable state of a structure and it is necessary to visualize the shape finding from unstable state to stable state. In this paper, a numerical method of stabilizing procedure for the link structures is presented. The structures are assumed to have rigid movements and thus only changing of the topology of member is considered during the analysis. The generalized inverse matrix and the principle of minimum potential energy are used in the process. Illustrative examples are presented and the results show good convergence.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.10 no.5
• /
• pp.611-617
• /
• 1986

In this paper, it is attempted to derive the minimum torque as the optimal value on each joint, which is applied during a PTP-motion in the range of working area of a supposed industrial robot. The rupposed industrial robot consits of 3-R joints prepared on three links, The optimizational analysis is performed by the formulation of a variational calculus process due to Rayleigh-Ritz method. That is, the torques of the inverse dynamic problem on joints in a arbitrary positions are computed by a generalized inertia matrix method.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.20 no.9
• /
• pp.2800-2811
• /
• 1996

The deformation of a structure shall be called homologous, if a given geometrical relation holds, for a given number of structural points, before, during, and after the deformation. Some researchers have utilized the idea on structural design with finite element method. The approaches use the decomposition of the FEM equation or equality of eqality equations to obtain homologous deformation. However, weight reduction and response constraints such as stress, displacement or natural frequency cannot be considered by those theories. An optimization method solving the above problems is suggested to gain homologous deformation. Homology constraints can be considered under multiple loadindg conditions as well as a single loading condition. Homology index is defined for the multiple loading conditions Examples are solved to present the performances of the method.