• International Journal of Advanced Culture Technology
• /
• 제7권4호
• /
• pp.56-62
• /
• 2019

Recently, research using artificial neural networks has further expanded the field of neural network optimization and automatic structuring from improving inference accuracy. The performance of the machine learning algorithm depends on how the hyperparameters are configured. Open-source hyperparameter optimization software can be an important step forward in improving the performance of machine learning algorithms. In this paper, we review open-source hyperparameter optimization softwares.

• Journal of the Korean Data and Information Science Society
• /
• 제17권3호
• /
• pp.905-912
• /
• 2006

Support vector classification(SVC) provides more complete description of the linear and nonlinear relationships between input vectors and classifiers. In this paper we propose to solve the optimization problem of SVC with a modified hinge loss function, which enables to use an iterative reweighted least squares(IRWLS) procedure. We also introduce the approximate cross validation function to select the hyperparameters which affect the performance of SVC. Experimental results are then presented which illustrate the performance of the proposed procedure for classification.

• Journal of the Korean Data and Information Science Society
• /
• 제18권3호
• /
• pp.767-772
• /
• 2007

A kernel machine is proposed as an estimating procedure for the linear and nonlinear Poisson regression, which is based on the penalized negative log-likelihood. The proposed kernel machine provides the estimate of the mean function of the response variable, where the canonical parameter is related to the input vector in a nonlinear form. The generalized cross validation(GCV) function of MSE-type is introduced to determine hyperparameters which affect the performance of the machine. Experimental results are then presented which indicate the performance of the proposed machine.

• Journal of the Korean Statistical Society
• /
• 제32권2호
• /
• pp.121-150
• /
• 2003

The National Health Interview Survey (NHIS) is one of the surveys used to assess the health status of the US population. One indicator of the nation's health is the total number of doctor visits made by the household members in the past year, There is a substantial nonresponse among the sampled households, and the main issue we address here is that the nonrespones mechanism should not be ignored because respondents and nonrespondents differ. It is standard practice to summarize the number of doctor visits by the binary variable of no doctor visit versus at least one doctor visit by a household for each of the fifty states and the District of Columbia. We consider a nonignorable nonresponse model that expresses uncertainty about ignorability through the ratio of odds of a household doctor visit among respondents to the odds of doctor visit among all households. This is a hierarchical model in which a nonignorable nonresponse model is centered on an ignorable nonresponse model. Another feature of this model is that it permits us to "borrow strength" across states as in small area estimation; this helps because some of the parameters are weakly identified. However, for simplicity we assume that the hyperparameters are fixed but unknown, and these hyperparameters are estimated by the EM algorithm; thereby making our method Bayes empirical Bayes. Our main result is that for some of the states the nonresponse mechanism can be considered non-ignorable, and that 95% credible intervals of the probability of a household doctor visit and the probability that a household responds shed important light on the NHIS.

• Asian-Australasian Journal of Animal Sciences
• /
• 제31권12호
• /
• pp.1863-1870
• /
• 2018

Objective: The Bayesian first-order antedependence models, which specified single nucleotide polymorphisms (SNP) effects as being spatially correlated in the conventional BayesA/B, had more accurate genomic prediction than their corresponding classical counterparts. Given advantages of $BayesC{\pi}$ over BayesA/B, we have developed hyper-$BayesC{\pi}$, ante-$BayesC{\pi}$, and ante-hyper-$BayesC{\pi}$ to evaluate influences of the antedependence model and hyperparameters for $v_g$ and $s_g^2$ on $BayesC{\pi}$.Methods: Three public data (two simulated data and one mouse data) were used to validate our proposed methods. Genomic prediction performance of proposed methods was compared to traditional $BayesC{\pi}$, ante-BayesA and ante-BayesB. Results: Through both simulation and real data analyses, we found that hyper-$BayesC{\pi}$, ante-$BayesC{\pi}$ and ante-hyper-$BayesC{\pi}$ were comparable with $BayesC{\pi}$, ante-BayesB, and ante-BayesA regarding the prediction accuracy and bias, except the situation in which ante-BayesB performed significantly worse when using a few SNPs and ${\pi}=0.95$. Conclusion: Hyper-$BayesC{\pi}$ is recommended because it avoids pre-estimated total genetic variance of a trait compared with $BayesC{\pi}$ and shortens computing time compared with ante-BayesB. Although the antedependence model in $BayesC{\pi}$ did not show the advantages in our study, larger real data with high density chip may be used to validate it again in the future.

• 한국IT서비스학회지
• /
• 제21권4호
• /
• pp.91-103
• /
• 2022

This study proposes a model for predicting contracted power using electric power data collected in real time from convenience stores nationwide. By optimizing the prediction model using machine learning, it will be possible to predict the contracted power required to renew the contract of the existing convenience store. Contracted power is predicted through the XGBoost regression model. For the learning of XGBoost model, the electric power data collected for 16 months through a real-time monitoring system for convenience stores nationwide were used. The hyperparameters of the XGBoost model were tuned using the GridesearchCV, and the main features of the prediction model were identified using the xgb.importance function. In addition, it was also confirmed whether the preprocessing method of missing values and outliers affects the prediction of reduced power. As a result of hyperparameter tuning, an optimal model with improved predictive performance was obtained. It was found that the features of power.2020.09, power.2021.02, area, and operating time had an effect on the prediction of contracted power. As a result of the analysis, it was found that the preprocessing policy of missing values and outliers did not affect the prediction result. The proposed XGBoost regression model showed high predictive performance for contract power. Even if the preprocessing method for missing values and outliers was changed, there was no significant difference in the prediction results through hyperparameters tuning.

• 한국방사선학회논문지
• /
• 제18권1호
• /
• pp.21-26
• /
• 2024

You only look once v5 (YOLOv5)는 객체 검출 과정에 우수한 성능을 보이고 있는 딥러닝 모델 중 하나다. 그러므로 본 연구의 목적은 양전차방출단층촬영 팬텀 영상에서 다양한 하이퍼 파라미터에 따른 YOLOv5 모델의 성능을 평가했다. 데이터 세트는 500장의 QIN PET segmentation challenge로부터 제공되는 오픈 소스를 사용하였으며, LabelImg 소프트웨어를 사용하여 경계박스를 설정했다. 학습의 적용된 하이퍼파라미터는 최적화 함수 SDG, Adam, AdamW, 활성화 함수 SiLu, LeakyRelu, Mish, Hardwish와 YOLOv5 모델 크기에 따라 nano, small, large, xlarge다. 학습성능을 평가하기 위한 정량적 분석방법으로 Intersection of union (IOU)를 사용하였다. 결과적으로, AdmaW의 최적화 함수, Hardwish의 활성화 함수, nano 크기에서 우수한 객체 검출성능을 보였다. 결론적으로 핵의학 영상에서의 객체 검출 성능에 대한 YOLOV5 모델의 유용성을 확인하였다.

• Communications for Statistical Applications and Methods
• /
• 제15권6호
• /
• pp.1003-1011
• /
• 2008

Mixed-effect Poisson regression models are widely used for analysis of correlated count data such as those found in longitudinal studies. In this paper, we consider kernel extensions with semiparametric fixed effects and parametric random effects. The estimation is through the penalized likelihood method based on kernel trick and our focus is on the efficient computation and the effective hyperparameter selection. For the selection of hyperparameters, cross-validation techniques are employed. Examples illustrating usage and features of the proposed method are provided.

• Communications for Statistical Applications and Methods
• /
• 제6권2호
• /
• pp.523-532
• /
• 1999

Generalized linear models have various applications for data arising from many kinds of statistical studies. Although the response variable is generally assumed to be generated from a wide class of probability distributions we focus on count data that are most often analyzed using binomial models for proportions or poisson models for rates. The methods and results presented here also apply to many other categorical data models in general due to the relationship between multinomial and poisson sampling. The novelty of the approach suggested here is that all conditional distribution s can be specified directly so that staraightforward Gibbs sampling is possible. The prior distribution consists of two stages. We rely on a normal nonconjugate prior at the first stage and a vague prior for hyperparameters at the second stage. The methods are demonstrated with an illustrative example using data collected by Rosenkranz and raftery(1994) concerning the number of hospital admissions due to back pain in Washington state.

• Journal of the Korean Data and Information Science Society
• /
• 제26권1호
• /
• pp.209-216
• /
• 2015

Quantile regression models with covariate measurement errors have received a great deal of attention in both the theoretical and the applied statistical literature. A lot of effort has been devoted to develop effective estimation methods for such quantile regression models. In this paper we propose the partially linear support vector orthogonal quantile regression model in the presence of covariate measurement errors. We also provide a generalized approximate cross-validation method for choosing the hyperparameters and the ratios of the error variances which affect the performance of the proposed model. The proposed model is evaluated through simulations.