• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.66 no.12
• /
• pp.1772-1781
• /
• 2017

In this study, the design methodology for alleviating the overfitting problem of Polynomial Neural Networks(PNN) is realized with the aid of two kinds techniques such as L2 regularization and Sum of Squared Coefficients (SSC). The PNN is widely used as a kind of mathematical modeling methods such as the identification of linear system by input/output data and the regression analysis modeling method for prediction problem. PNN is an algorithm that obtains preferred network structure by generating consecutive layers as well as nodes by using a multivariate polynomial subexpression. It has much fewer nodes and more flexible adaptability than existing neural network algorithms. However, such algorithms lead to overfitting problems due to noise sensitivity as well as excessive trainning while generation of successive network layers. To alleviate such overfitting problem and also effectively design its ensuing deep network structure, two techniques are introduced. That is we use the two techniques of both SSC(Sum of Squared Coefficients) and $L_2$ regularization for consecutive generation of each layer's nodes as well as each layer in order to construct the deep PNN structure. The technique of $L_2$ regularization is used for the minimum coefficient estimation by adding penalty term to cost function. $L_2$ regularization is a kind of representative methods of reducing the influence of noise by flattening the solution space and also lessening coefficient size. The technique for the SSC is implemented for the minimization of Sum of Squared Coefficients of polynomial instead of using the square of errors. In the sequel, the overfitting problem of the deep PNN structure is stabilized by the proposed method. This study leads to the possibility of deep network structure design as well as big data processing and also the superiority of the network performance through experiments is shown.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 1989.04a
• /
• pp.17-21
• /
• 1989

SSI2D/3D is a computer program to calculate dynamic stiffness matrix of the foundation for soil-structure-interaction problem in frequency demain. It is written in FORTRAN 77 and applicable to two or three dimensional situations. In this paper the program structure is summarized. Two examples aye shown to demonstrate the possibilities of the Boundary Element Method applied to dynamic problems in infinite domains.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2000.10a
• /
• pp.443-448
• /
• 2000

In this paper, multiphase optimization of machine structure is presented. The goal of first step is to obtain (i) light weight, (ii) rigidity statically. In this step, multiple optimization problem with two objective functions is treated using Pareto Genetic Algorithm. Where two objective functions are weight of the structure, and static compliance. The method is applied to a new machine structure design.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2011.06a
• /
• pp.187-188
• /
• 2011

Independent component analysis (ICA) is a method separating the mixture of signals into statistically and mutually independent ones. It has been applied to not only the Cocktail-party problem but also EEG analysis using the EEG waveform, digital signal processing, image processing and cognitive technique field actively. This study aims to propose a procedure to estimate the modal responses and mode shapes of a floating structure by using the ICA method from measured responses of the floating structure.

• Journal of the Korean Professional Engineers Association
• /
• v.19 no.2
• /
• pp.21-25
• /
• 1986

Structure-borne noise due to forced vibration which is originated from subway operation and transmitted to buildings in order of rail-wood tie-concrete bed-structure-soil-building foundations-members of building results in social problem of environmental pollution. Moreover this becomes a serious problem because of the increment of surface traffic and subway operation made by meeting traffic system in crowded cities. Since subway is constructed along the principal road or through the residential area and as the worst case may be, building foundations is contact with top part of subway structure, it is possible that vibration resonance results in fatal damage of buildings. And, structure-borne vibration noise due to subway operation at late and early hours have the residents suffer from insonmia, restlessness and so on. Therefore, to satisfy the future need concerning the environmental protection, this report deals with the influence of structure-borne vibration noise on the basis of the characteristics of Seoul Subway System.

• The Transactions of the Korean Institute of Electrical Engineers
• /
• v.35 no.6
• /
• pp.246-256
• /
• 1986

Determination of an aerodynamic structure is a very important problem in missile modeling. The structure problem is to choose an appropriate set of aerodynamic coefficients to represent chosen missile dynamics. A methodology and criteria to determine a structure from windtunnel data are presented in this paper. Aerodynamic coeffecients in the determined structure are then identified by parameter identification algorithms. The identified coefficients are in turn used to verify appropriateness of the structure. The extended Kalman filter (EKF) and the maximum likelihood mithod (ML) are adopted as the parameter identification algorithm. Both methods exhibit satisfactory results. While the model identified by the ML more closely follows dynamics of the chosen missile than that by the EKF.

• Coupled systems mechanics
• /
• v.1 no.1
• /
• pp.99-113
• /
• 2012

We present a monolithic semi-implicit algorithm for solving fluid-structure interaction problem at small structural displacements. The algorithm uses one global mesh for the fluid-structure domain obtained by gluing the fluid and structure meshes which are matching on the interface. The continuity of velocity at the interface is automatically satisfied and the continuity of stress does not appear explicitly in the global weak form due to the action and reaction principle. At each time step, we have to solve a monolithic system of unknowns velocity and pressure defined on the global fluid-structure domain. Numerical results are presented.

• The Mathematical Education
• /
• v.45 no.1 s.112
• /
• pp.123-138
• /
• 2006

In this paper, we confirm through surveys and interviews that it helps students in solving a problem counting the number of combinations to find a structural isomorphism between the given problem and a typical problem with the same mathematical structure. Then we suggest that a problem of distributing balls into boxes might be a good candidate for a typical problem. This approach is coherent to the viewpoint given by English(2004) that it is educationally important to see the connection and relationship between problems with different context but with similar mathematical structure.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 1996.04a
• /
• pp.182-189
• /
• 1996

The focus of this study is on the problem of the design of structure of undetermined topology. This problem has been regarded as being the most challenging of structural optimization problems, because of the difficulty of allowing topology to change. Conventional approaches break down when element sizes approach to zero, due to stiffness matrix singularity. In this study, a novel nonlinear Programming formulation of the topology Problem is developed and examined. Its main feature is the ability to account for topology variation through zero element sizes. Stiffness matrix singularity is avoided by embedding the equilibrium equations as equality constraints in the optimization problem. Although the formulation is general, two dimensional plane elasticity examples are presented. The design problem is to find minimum weight of a plane structure of fixed geometry but variable topology, subject to constraints on stress and displacement. Variables are thicknesses of finite elements, and are permitted to assume zero sizes. The examples demonstrate that the formulation is effective for finding at least a locally minimal weight.

• 한국보건간호학회:학술대회논문집
• /
• 2006.06a
• /
• pp.30-43
• /
• 2006