• Communications for Statistical Applications and Methods
• /
• 제23권4호
• /
• pp.355-362
• /
• 2016

Ensemble methods often help increase prediction ability in various predictive models by combining multiple weak learners and reducing the variability of the final predictive model. In this work, we demonstrate that ensemble methods also enhance the accuracy of prediction under kernel ridge regression and kernel logistic regression classification. Here we apply bagging and random forests to two kernel-based predictive models; and present the procedure of how bagging and random forests can be embedded in kernel-based predictive models. Our proposals are tested under numerous synthetic and real datasets; subsequently, they are compared with plain kernel-based predictive models and their subsampling approach. Numerical studies demonstrate that ensemble approach outperforms plain kernel-based predictive models.

• 산업공학
• /
• 제23권3호
• /
• pp.229-238
• /
• 2010

Virtual metrology (VM), a critical component of semiconductor manufacturing, is an efficient way of assessing the quality of wafers not actually measured. This is done based on a model between equipment sensor data (obtained for all wafers) and the quality characteristics of wafers actually measured. This paper considers principal component regression (PCR), partial least squares regression (PLSR), kernel PCR (KPCR), and kernel PLSR (KPLSR) as VM models. For each regression model, two cases are considered. One utilizes all explanatory variables in developing a model, and the other selects significant variables using the genetic algorithm (GA). The prediction performances of 8 regression models are compared for the short- and long-term etch process data. It is found among others that the GA-KPLSR model performs best for both types of data. Especially, its prediction ability is within the requirement for the short-term data implying that it can be used to implement VM for real etch processes.

• Computers and Concrete
• /
• 제24권1호
• /
• pp.7-17
• /
• 2019

The complex phenomenon of the bond formation in geopolymer is not well understood and therefore, difficult to model. This paper present applied statistical models for evaluating the compressive strength of geopolymer. The applied statistical models studied are divided into three different categories - linear regression [least absolute shrinkage and selection operator (LASSO) and elastic net], tree regression [decision and bagging tree] and kernel methods (support vector regression (SVR), kernel ridge regression (KRR), Gaussian process regression (GPR), relevance vector machine (RVM)]. The performance of the methods is compared in terms of error indices, computational effort, convergence and residuals. Based on the present study, kernel based methods (GPR and KRR) are recommended for evaluating compressive strength of Geopolymer concrete.

• Journal of the Korean Data and Information Science Society
• /
• 제20권5호
• /
• pp.949-954
• /
• 2009

The autoregressive process is applied in this paper to kernel regression in order to infer nonlinear models for predicting responses. We propose a kernel method for the autoregressive data which estimates the mean function by kernel machines. We also present the model selection method which employs the cross validation techniques for choosing the hyper-parameters which affect the performance of kernel regression. Artificial and real examples are provided to indicate the usefulness of the proposed method for the estimation of mean function in the presence of autocorrelation between data.

• 대한음성학회지:말소리
• /
• 제66호
• /
• pp.105-116
• /
• 2008

In this research, we propose a speaker identification system using a kernel method which is expected to model the non-linearity of speech features well. We have been using principal component analysis (PCA) successfully, and extended to kernel PCA, which is used for many pattern recognition tasks such as face recognition. However, we cannot use kernel PCA for speaker identification directly because the storage required for the kernel matrix grows quadratically, and the computational cost grows linearly (computing eigenvector of $l{\times}l$ matrix) with the number of training vectors I. Therefore, we use greedy kernel PCA which can approximate kernel PCA with small representation error. In the experiments, we compare the accuracy of the greedy kernel PCA with the baseline Gaussian mixture models using MFCCs and PCA. As the results with limited enrollment data show, the greedy kernel PCA outperforms conventional methods.

• Journal of the Korean Data and Information Science Society
• /
• 제21권3호
• /
• pp.419-425
• /
• 2010

커널기계 기법은 최근 대용량 또는 고차원 비선형 자료를 분석하는 방법으로 인기를 많이 얻고 있다. 본 논문에서는 주식시장 수익률의 조건부 변동성을 예측하기 위한 일반화 이분산자기회귀모형을 추정하기 위해 커널기계 기법을 사용한다. 일반화 이분산자기회귀모형은 자료가 정규분포를 따른다고 가정한 후 주로 최대우도법을 사용하여 추정된다. 본 논문에서는 꼬리가 두꺼운 분포를 갖는 금융시계열자료의 변동성을 추정할 때 커널기계 기법이 최대우도법과 서포트벡터기계 보다 더 정확한 예측능력을 가진다는 것을 보이고자 한다.

• 대한수학회지
• /
• 제44권4호
• /
• pp.835-844
• /
• 2007

This paper considers the problem of estimating the error distribution function in nonparametric regression models. Sufficient conditions are given under which the kernel estimator of the error distribution function based on nonparametric residuals satisfies the law of iterated logarithm.

• 한국통신학회논문지
• /
• 제36권1A호
• /
• pp.35-40
• /
• 2011

랜덤 발생 심볼과 출력 신호에 대해 두 확률 밀도 함수 사이의 유클리드 거리를 최소화하는 복소 채널 등화 알고리듬은 정보 이론적 학습방법의 장점을 살리면서 위상 회전 문제까지 극복할 수 있도록 설계 되었다. 이 논문에서는 이 알고리듬에 대해 확률 밀도 함수 구축에 사용된 커널 사이즈가 성능에 끼치는 영향을 연구하였고 커널 사이즈의 변형에 인한 정보 포텐셜 간의 힘 조절에 변화를 준 Kernel-modified 알고리듬을 제안하였다. 이 제안한 방식은 커널 사이즈 변형이 이루어지지 않은 알고리듬에 대해 약 4 dB 정도의 성능 향상을 만들어 냈다. 성상도 특성에서도 복소 채널에 의한 위상 회전이 완벽하게 극복될 뿐 아니라 보다 집중된 심볼 점을 보였다.

• 한국방재안전학회논문집
• /
• 제8권1호
• /
• pp.15-19
• /
• 2015

수문학에서 사용되는 강우-유출 모델의 경우 선형적인 시스템을 기반으로 유효강수량으로부터 시간적 지연을 통해서 유출량이 결정되는데 그 양은 강우량의 선형적인 비로 표현되어서 결국 합성곱을 통해 해석되게 된다. 또한 자료에 대한 확률론적 분석에 많이 이용되는 비매개변수 핵밀도함수의 경우, 핵(Kernel)의 의미자체가 합성곱에서 나온 것으로서 개개의 자료를 바탕으로 핵을 통해 매끄러운 확률밀도함수를 구하게 된다. 본 연구에서는 합성곱을 바탕으로 강우-유출 모델과 비매개변수 확률밀도함수를 해석하는 방법에 대해서 되짚어 보고 그 공통적인 특성과 다른 점을 수학적으로 나타내 줌으로써 사용되는 합성곱 함수의 유용성에 대해서 논하였다.

• Journal of Biosystems Engineering
• /
• 제23권1호
• /
• pp.49-56
• /
• 1998

The objectives of this study were to develop models to predict the contents of moisture and protein of single kernel of brown rice based on visible/NIR (near-infrared) spectroscopic technique. The reflectance spectra of rice were obtained in the range of the wavelength 400 to 2,500 nm with 2 nm intervals. Multiple linear regression(MLR) and partial least squares (PLS) were used to develop the models. The MLR model using the first derivative spectra(10 nm of gap) with Standard Normal Variate and Detrending (SNV and Drt.) preprocessing showed the best results to predict moisture content of the sin린e kernel brown rice. To predict the protein content of a single kernel of brown ricer the PLS model used the raw spectra with multiplicative scatter correction(MSC) preprocessing over the wavelength of 1,100~1,500 nm.