• Proceedings of the Korean Statistical Society Conference
• /
• 2000.11a
• /
• pp.219-225
• /
• 2000

데이터 마이닝 분야에서 널리 사용되고 있는 신경망은 최근 많은 통계인들의 관심을 끌고 있다. 그러나 범용 근사성(universal approximator)이라는 성질에도 불구하고 초기치에 따라 적합 결과가 크게 좌우되는 단점이 있다. 본 논문에서는 붓스트랩 표본을 통해 초기치를 발견하는 bumping 기법이 신경망 분야에서 사용되고 있는 무작위 탐색법 보다 더 정확하고 안정적인 초기치를 제공하여 주는가를 살펴 보았다.

• The Korean Journal of Applied Statistics
• /
• v.18 no.2
• /
• pp.411-420
• /
• 2005

In this study, the MIVQUE estimators of variance components for the one-way random model with unbalanced data are investigated. In order to compare the efficiency of MIVQUE estimators obtained by using three priori estimates, the Empirical Quantile Dispersion Graphs (EQDGs) are used. From the results of Monte-Carlo study, the MIVQUE estimator using ${\sigma}^2_{\alpha}\;=\;0\;and\;{\sigma}^2_{varraho}=1$ as the priori estimate performs well relative to other estimators.

• Journal of IKEEE
• /
• v.16 no.3
• /
• pp.211-216
• /
• 2012

In the phase boundary estimation problem, the estimation performance depends on the initial guess. However, there is no information on the number of bubbles and those positions for the initial guess in real flows. Therefore, it is very important to set appropriate initial guesses from prior information. In this paper, in order to set initial guesses for estimating the phase boundaries in two-phase flows, first, unknown resistivity distribution was estimated using the difference reconstruction method. After that, an adaptive threshold value was automatically computed using intermodes method. Based on this value, the number of bubbles and the initial position were determined. The numerical experiments have been performed to evaluate the estimation performance of the proposed method.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.13 no.4
• /
• pp.416-426
• /
• 2003

In this paper we propose a new method for choosing the initial values of Expectation-Maximization(EM) algorithm that has been used in various applications for clustering. Conventionally, the initial values were chosen randomly, which sometimes yields undesired local convergence. Later, K-means clustering method was employed to choose better initial values, which is currently widely used. However the method using K-means still has the same problem of converging to local points. In order to resolve this problem, a new method of initializing values for the EM process. The proposed method not only strengthens the characteristics of EM such that the number of iteration is reduced in great amount but also removes the possibility of falling into local convergence.

• Proceedings of the KIEE Conference
• /
• 2007.07a
• /
• pp.1652-1653
• /
• 2007

가변제어구조로 동작하는 슬라이딩모드제어(SMC)는 플렌트의 파라미터 변동과 부하왜형에 관계없이 스위칭 제어 알고리즘에 의해 위상 평면에서 미리 예측된 궤적 또는 기준모델을 따라 구동응답을 주어진 슬라이딩 면을 따라 강제로 추종시키는 것이다. 여기서 슬라이딩 평면을 찾아내는 방법의 하나로 SVM(Suppot Vector Machine) 을 사용한다. 그런데 SVM을 사용하여 슬라이딩 평면을 찾아내는 과정에서, 초기치의 변동이 있을 경우, SVM 모델을 재구성해야 해야 한다. 이에 본 논문에서는 SVM 모델을 재구성할 필요 없이, 기존 초기치에 의한 SVM 모델에서, 원하는 초기치의 SVM모델로 변경할 수 있는 방법을 제안한다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.7 no.2
• /
• pp.284-292
• /
• 2003

In this paper, we carried out the numerical examination of the initial value setting-up method to extend the range of optimal step parameter in a adaptive system which is controlled by LMS algorithm. For initial value setting-up methods, the general method which select the initial value randomly and the other method which applies the approximate value obtained from the direct method to initial value, were used. And then, we compared to the ranges of step parameter setting, the convergence speeds of mean-square-error, and the stabilities during the convergence process when the initial values were applied to the optimal directivity synthesis problem. According to the numerical simulation results, the initial value setting-up method by means of the direct method provides wider range for the step parameter, more efficient capability for convergence and stability, and more error correction ability than the general method.

• Proceedings of the Korean Statistical Society Conference
• /
• 2003.05a
• /
• pp.19-24
• /
• 2003

신경망은 적용 다양성과 제약조건의 최소성, 강력한 예측성, 범용성, 근사성 등 많은 장점을 지니고 있으나 초기 가중치의 할당에 따라 모델 생성의 Performance와 예측의 결과가 달라지게 되는 단점을 지니고 있다. 이런 신경망의 초기 가중치에 따른 단점을 보안하기 위해 통계적 알고리즘의 접목을 통해 Hybrid된 신경망 보완 알고리즘을 제시하고자 하였다. 논문을 위한 기본 가정으로 신경망의 가장 기본인 SLP 알고리즘을 바탕으로 활성함수에 가장 일반적으로 사용되는 Sigmoid 활성함수를 이용하였을 때, 초기 가중치로 기존의 임의 난수 생성 방식이 아닌 통계적 로지스틱 회귀분석의 계수값(mle)을 제시하여 이를 초기치로 사용한 경우와 그렇지 않은 경우의 예측 정확성과 수렴의 Performance정도를 비교하여 가장 효과적인 초기치 방법을 제시하고자 하였다.

• The Journal of the Korea Contents Association
• /
• v.19 no.1
• /
• pp.648-659
• /
• 2019

The social network services have spread rapidly with new issues in terms of the security. The purpose of this study is to longitudinally verify the influence of relationships among online security, social network service, and smartphone expense. The analyses were conducted on panel data of SNS users from 2015 to 2017 based on the SEM methodologies. The results indicated that anxiety of online security had an effect on the SNS usage. In addition, The results showed that anxiety of online security and SNS usage had a significant influence on the smartphone expenses. This study could provide useful guidelines for researchers and company managers, and an opportunity to develop an understanding of social network trends.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.18 no.2
• /
• pp.241-250
• /
• 2023

In this paper, when using the LMS algorithm for adaptive ultrasonic beamforming system, the effect of initial value setting on the margin of step parameters was studied. To this end, quasi-ideal beams, rotational beams with arbitrarily set beam widths were used as examples. In the numerical simulations, an arbitrary initial value(the number of sound sources fixed to any number) was set in the ultrasonic beamforming system, and the margin of the step parameter and convergence characteristics thereof were compared.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.20 no.4
• /
• pp.698-706
• /
• 2016

In order to move mobile robots to desired locations in a minimum time, optimal control problems have to be solved; however, their analytic solutions are almost impossible to obtain due to robot nonlinear equations. This paper presents a method to get optimal control gains of mobile robots using genetic algorithms. Since the optimal control gains of mobile robots depend on the initial conditions, the initial condition range is discretized to form some grid points, and genetic algorithms are applied to provide the optimal control gains for the corresponding grid points. The optimal control gains for general initial conditions may be obtained by use of neural networks. So the optimal control gains and the corresponding grid points are used to train neural networks. The trained neural networks can supply pseudo-optimal control gains. Finally simulation studies have been conducted to verify the effectiveness of the method presented in this paper.