• Smart Structures and Systems
• /
• v.6 no.1
• /
• pp.17-27
• /
• 2010

The approach to damage detection and localization adopted in this paper is based on a statistical comparison of models built from the response time histories collected at different stages during the structure lifetime. Some of these time histories are known to have been recorded when the structural system was undamaged. The consistency of the models associated to two different stages, both undamaged, is first recognized. By contrast, the method detects the discrepancies between the models from measurements collected for a damaged situation and for the undamaged reference situation. The damage detection and localization is pursued by a comparison of the SSE (sum of the squared errors) histograms. The validity of the proposed approach is tested by applying it to the analytical benchmark problem developed by the ASCE Task Group on Structural Health Monitoring (SHM). In the paper, the results of the benchmark studies are presented and the performance of the method is discussed.

• The Korean Journal of Applied Statistics
• /
• v.23 no.4
• /
• pp.759-765
• /
• 2010

In this paper, two new inverse regression methods are introduced. One is a distance based method, and the other is a likelihood based method. While a model is fitted by minimizing the sum of squared prediction errors of y's and x's in the classical and inverse methods, respectively. In the new distance based method, we simultaneously minimize the sum of both squared prediction errors. In the likelihood based method, we propose an inverse regression with Arnold-Beaver Skew Normal(ABSN) error distribution. Using the cross validation method with an asymmetric real data set, two new and two existing methods are studied based on the relative prediction bias(RBP) criteria.

• The KIPS Transactions:PartD
• /
• v.11D no.6
• /
• pp.1269-1276
• /
• 2004

The Parameter Estimation for software existing reliability models, Goel-Okumoto, Yamada-Ohba-Osaki model was reviewed and Rayleigh model based on Rayleigh distribution was studied. In this paper, we discusses comparison of parameter estimation using maximum likelihood estimator and Bayesian estimation based on Gibbs sampling to analysis of the estimator' pattern. Model selection based on sum of the squared errors and Braun statistic, for the sake of efficient model, was employed. A numerical example was illustrated using real data. The current areas and models of Superposition, mixture for future development are also employed.

• Communications for Statistical Applications and Methods
• /
• v.18 no.6
• /
• pp.733-739
• /
• 2011

The linear absolute shrinkage and selection operator(Lasso) method improves the low prediction accuracy and poor interpretation of the ordinary least squares(OLS) estimate through the use of $L_1$ regularization on the regression coefficients. However, the Lasso is not robust to outliers, because the Lasso method minimizes the sum of squared residual errors. Even though the least absolute deviation(LAD) estimator is an alternative to the OLS estimate, it is sensitive to leverage points. We propose a robust Lasso estimator that is not sensitive to outliers, heavy-tailed errors or leverage points.

• 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.

• Journal of Korean Society for Quality Management
• /
• v.33 no.1
• /
• pp.73-82
• /
• 2005

This article proposes a practical approach, which is based on the concept of the expected squared relative error, that can consider both the prediction quality and the practitioner's subjectivity in simultaneously optimizing multiple responses. Through a case study, multiresponse optimization using the expected squared relative error approach is illustrated, and the SAS program to implement the proposed method is provided.

• The Journal of the Acoustical Society of Korea
• /
• v.18 no.2E
• /
• pp.3-9
• /
• 1999

In this paper we study the convergence behavior of the least mean fourth(LMF) algorithm where the error raised to the power of four is minimized for a multiple sinusoidal input and Gaussian measurement noise. Here we newly obtain the convergence equation for the sum of the mean of the squared weight errors, which indicates that the transient behavior can differ depending on the relative sizes of the Gaussian noise and the convergence constant. It should be noted that no similar results can be expected from the previous analysis by Walach and Widrow/sup [1]/.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.26 no.12A
• /
• pp.2043-2049
• /
• 2001

In this paper we study the convergence behavior of the least mean fourth(LMF) algorithm where the error raised to the power of four is minimized for a multiple sinusoidal input and Gaussian measurement noise. Here we newly obtain the convergence equation for the sum of the mean of the squared weight errors, which indicates that the transient behavior can differ depending on the relative sizes of the Gaussian noise and the convergence constant. It should be noted that no similar results can be expected from the previous analysis by Walach add Widrow.

• Journal of the military operations research society of Korea
• /
• v.3 no.1
• /
• pp.123-136
• /
• 1977

This thesis presents a modification and extension to the Burgess and Killebrew heuristic resource leveling procedure for project networks. In contrast to previous algorithms appearing in the literature, the objective function of this algorithm. is the minimization of the sum of the squared errors in each time period (deviations around the mean usage) of all resources over the duration of the project. This objective function continues the search for an improved schedule beyond that of previous algorithms with their associated objective functions. One important feature is that the algorithm tends to reduce the number of periods that a resource is idle during its duration on the project.

• Proceedings of the IEEK Conference
• /
• 2001.09a
• /
• pp.915-918
• /
• 2001

In this paper we study a new convergence behavior of the least mean fourth (LMF) algorithm where the error raised to the power of four is minimized for a multiple sinusoidal input and Gaussian measurement noise. Here we newly obtain the convergence equation for the sum of the mean of the squared weight errors, which indicates that the transient behavior can differ depending on the relative sizes of the Gaussian noise and the convergence constant. It should be noted that no similar results can be expected from the previous analysis by Walach and Widrow.