• Communications for Statistical Applications and Methods
• /
• 제17권6호
• /
• pp.803-810
• /
• 2010

The singular value decomposition of a rectangular matrix is a basic tool to understand the structure of the data and particularly the relationship between row and column factors. However, conventional singular value decomposition used the least squares method and is not robust to outliers. We propose a simple robust singular value decomposition algorithm based on the weighted least absolute deviation which is not sensitive to leverage points. Its implementation is easy and the computation time is reasonably low. Numerical results give the data structure and the outlying information.

• 대한전기학회논문지
• /
• 제41권9호
• /
• pp.1021-1030
• /
• 1992

When using a least squares estimator with exponential forgetting factor to identify continuous-time deterministic system, the problem of determining minimum eigenvalue is described in this paper. It is well known fact that the convergence rate of parameter estimates relies on various factors consisting of the estimator and especially, theirproperties can be directly affected by all eigenvalues in the parameter error differential equation. Fortunately, there exists only one adjusting eigenvalue in the given estimator and then, the parameter convergence rates depend on this minimum eigenvalue. In this note, a new result to determine the minimum eigenvalue is proposed. Under the assumption that the input has as many spectral lines as the number of parameter estimates, it can be proven that the minimum eigenvalue converges to a constant value, which is a function of the forgetting factor and the parameter estimates number.

• Interaction and multiscale mechanics
• /
• 제3권2호
• /
• pp.123-143
• /
• 2010

The essential idea of the element-free Galerkin method (EFG) is that moving least-squares (MLS) approximation are used for the trial and test functions with the variational principle (weak form). By using the weighted orthogonal basis function to construct the MLS interpolants, we derive the formulae for an improved element-free Galerkin (IEFG) method for solving three-dimensional problems in linear elasticity. There are fewer coefficients in improved moving least-squares (IMLS) approximation than in MLS approximation. Also fewer nodes are selected in the entire domain with the IEFG method than is the case with the conventional EFG method. In this paper, we selected a few example problems to demonstrate the applicability of the method.

• 한국정보통신학회논문지
• /
• 제24권2호
• /
• pp.219-224
• /
• 2020

이진 분류(binary classification)는 머신러닝(machine learning) 분야에서 많이 다루어진 주제이다. 게다가 이진 분류는 다중 분류로 쉽게 발전될 수 있는 중요한 분야이다. 머신러닝 방법들을 적용할 때에 전처리(preprocessing)이나 특징 추출(feature extraction)과 같은 작업이 필수적이다. 이는 분류기 성능을 향상시키기 위한 중요한 작업이다. 본 논문에서는 가중된 최소 자승법을 기반으로 새로운 머신러닝 방법을 제안한다. 또한, 특징 변환시킬 수 있는 새로운 가중치 계산 방법을 제안한다. 이를 통해 특징 변환과 동시에 학습을 진행할 수 있는 방법을 제안한다. 본 제안을 다섯 개의 머신러닝 데이터베이스에서 실험을 진행하였으며 이 데이터베이스에서 우수한 성능을 얻을 수 있었다.

• Journal of the Korean Data and Information Science Society
• /
• 제25권4호
• /
• pp.931-939
• /
• 2014

In this paper we propose the iteratively reweighted least squares procedure to solve the quadratic programming problem of support vector expectile regression with an asymmetrically weighted squares loss function. The proposed procedure enables us to select the appropriate hyperparameters easily by using the generalized cross validation function. Through numerical studies on the artificial and the real data sets we show the effectiveness of the proposed method on the estimation performances.

• Industrial Engineering and Management Systems
• /
• 제9권3호
• /
• pp.262-274
• /
• 2010

This article investigates how to adaptively predict the time-varying metrology delay that could realistically occur in the semiconductor manufacturing practice. Metrology delays pose a great challenge for the existing run-to-run (R2R) controllers, driving the process output significantly away from target if not adequately predicted. First, the expected asymptotic double exponentially weighted moving average (DEWMA) control output, by using the EWMA and recursive least squares (RLS) prediction methods, is derived. It has been found that the relationships between the expected control output and target in both estimation methods are parallel, and six cases are addressed. Within the context of time-varying metrology delay, this paper presents a modified recursive least squares-linear trend (RLS-LT) controller, in combination with runs test. Simulated single input-single output (SISO) R2R processes subject to various time-varying metrology delay scenarios are used as a testbed to evaluate the proposed algorithms. The simulation results indicate that the modified RLS-LT controller can yield the process output more accurately on target with smaller mean squared error (MSE) than the original RLSLT controller that only deals with constant metrology delays.

• 대한핵의학회지
• /
• 제37권5호
• /
• pp.288-300
• /
• 2003

목적: ${H_2}^{15}O$ PET의 정량화를 위하여 1-조직 구획모델이 쓰이며, 뇌혈류와 조직/혈액 분배계수를 구하기 위하여 nonlinear least squares (NLS) 방법이 사용되나 계산 시간이 긴 등의 문제로 파라미터를 각화소마다 구해야 하는 파라메트릭 영상 구성에는 적합하지 않다. 이 연구에서는 이와 같은 NLS 문제점을 극복하여 파라메트릭 영상을 빠르게 구성하기 위하여 제안된 파라미터 추정 알고리즘들을 구현하고, 이 방법들의 통계적 신뢰도와 계산의 효율성을 비교하였다. 대상 및 방법: 이 연구에서 이용한 방법들은 linear least squares (LLS), linear weighted least squares (LWLS), linear generalized least squares (GLS), linear generalized weighted least squares (GWLS), weighted integration (WI), 그리고 model-based clustering method (CAKS)이다. 노이즈 정도에 따른 각 파라메트릭 영상법의 정확성 및 통계적 신뢰성을 알아보기 위하여 Zubal 뇌모형(brain phantom)으로부터 동적 PET 영상을 모사하고 포아송노이즈를 더한 후 각 파라메트릭 영상 구성 방법을 적용하였다. 또한 정상인 16명에 대하여 얻은 실제 자료에 대하여 이 방법들을 적용하고 결과를 비교하였다. 결과: 뇌혈류와 분배계수에 대한 평균 오차는 방법에 따라 크게 다르지 않았으며 모든 방법이 뇌혈류 및 분배계수 추정에 있어 무시할 만한 바이어스를 보였다. 파라메트릭 영상의 정성적 특성 또한 유사하였으나 CAKS 방법의 계산 속도가 월등하여 NLS 방법의 약 1/500, LLS 방법의 약 1/25의 계산시간을 보였다. 결론: 뇌혈류 파라메트릭 영상 구성을 위한 빠른 파라미터 추정 알고리즘들 중에 보다 개선되어 제안된 LWS, GLS, GLWS, CAKS 방법들이 단순하고 빠른 LLS, WI 방법들에 비하여 통계적 신뢰성을 크게 향상시키지는 못하나 CAKS 방법은 계산 시간을 유의하게 단축시키므로 가장 적합한 파라메트릭 영상 구성방법이라 할 수 있을 것이다.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2007년도 제38회 하계학술대회
• /
• pp.1640-1641
• /
• 2007

In this paper, an experiment of weighted robust least squares frequency estimation for the Gaussian envelope chirp signal which is used in the time-frequency domain reflectometry system was carried out. By incorporating the forgetting factor to the frequency estimator, the weighted robust least squares filter achieved good enough frequency estimation performance for the chirp signal and it can be adopted to implement not only low cost time-frequency domain reflectometry but also real-time time-frequency domain reflectometry implementation.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1991년도 하계학술대회 논문집
• /
• pp.691-694
• /
• 1991

In a discrete parameter estimation system, the standard least squares method shows slow convergence. On the other hand, the weighted least squares method has relatively fast convergence. However, if the input is not sufficiently rich, then gain matrix grows unboundedly. In order to solve these problems, this paper proposes a modified least squares algorithm which prevents gain matrix from growing unboundedly and has fast convergence.

• Journal of the Korean Data and Information Science Society
• /
• 제17권2호
• /
• pp.401-409
• /
• 2006

We consider a generalized partly-parametric additive risk model which generalizes the partly parametric additive risk model suggested by McKeague and Sasieni (1994). As an estimation method of this model, we propose to use the weighted least square estimation, suggested by Huffer and McKeague (1991), for Aalen's additive risk model by a piecewise constant risk. We provide an illustrative example as well as a simulation study that compares the performance of our method with the ordinary least squares method.