• Journal of KIISE:Software and Applications
• /
• v.31 no.12
• /
• pp.1569-1580
• /
• 2004

One commonly used approach to deal with uncertainty is Bayesian network which represents joint probability distributions of domain. There are some attempts to team the structure of Bayesian networks automatically and recently many researchers design structures of Bayesian network using evolutionary algorithm. However, most of them use the only one fittest solution in the last generation. Because it is difficult to combine all the important factors into a single evaluation function, the best solution is often biased and less adaptive. In this paper, we present a method of generating diverse Bayesian network structures through fitness sharing and combining them by Bayesian method for adaptive inference. In order to evaluate performance, we conduct experiments on learning Bayesian networks with artificially generated data from ASIA and ALARM networks. According to the experiments with diverse conditions, the proposed method provides with better robustness and adaptation for handling uncertainty.

• The Bulletin of The Korean Astronomical Society
• /
• v.44 no.1
• /
• pp.85.1-85.1
• /
• 2019

Accurate estimation of the masses, the ages, and the chemical abundances of host stars is crucial to understand physical characteristics of exo-planetary systems. In this study, we investigate physical dimensions of 94 planet-hosting stars based on spectroscopic observation and stellar evolutionary computation, From the high resolution echelle spectroscopy of the BOES observation, we have analysed metallicities and alpha-element enhancements of host stars. By combining recent spectro-photometric observations, stellar parameters are calibrated within the frame work of the standard stellar theory. In general, the minimum chi-square estimation can be strongly biased in cases that stellar properties rapidly changes after the terminal age main-sequence. Instead, we adopt a Bayesian statistics considering a priori distribution of stellar parameters during the rapid evolutionary phases. we determine a reliable set of stellar parameters between theoretical model grids. To overcome this statistical bias, (1) we adopt a Bayesian statistics considering a priori distribution of stellar parameters during the rapid evolutionary phases and (2) we construct the fine model grid that covers mass range ($0.2{\sim}3.0M_{\odot}$) with the mass step ${\Delta}M=0.01M_{\odot}$, metallicities Z = 0.0001 ~ 0.04, and the helium and the alpha-element enhancement. In this presentation, we introduce our calibration scheme for several hosting stars.

• Proceedings of the Korean Information Science Society Conference
• /
• 2005.07b
• /
• pp.697-699
• /
• 2005

연속 변수 함수 최적화를 위한 진화 연산에서는 전통적으로 확률 분포를 도입하여 새로운 세대를 생성하는 기법을 사용하고 있다. 최근 들어 이러한 확률 분포를 개체군으로부터 추정하여 보다 효율적으로 최적화를 해결하려는 연구가 진행되고 있다. 본 논문에서는 variational 베이지안 혼합 인자 분석 기법(Bayesian mixtures of factor analyzers)을 사용한 개체군의 분포 추정을 통해 연속 변수 함수의 최적화 문제를 해결하는 방법을 제안한다. 이 기법은 혼합 분포의 개수 추정을 자동화하여 개체군의 다양성을 유지할 수 있기 때문에 지역 최적점으로 일찍 수렴하는 현상을 방지할 수 있으며, 세부 개체군 내의 분포 추정을 통해 탐색을 효율적으로 수행할 수 있다. 잘 알려진 평가 함수들에 대하여 다른 분포 추정 진화 연산과 비교하여 제안하는 방법의 우수성을 검증하였다.

• Journal of KIISE:Software and Applications
• /
• v.36 no.3
• /
• pp.208-216
• /
• 2009

A number of estimation of distribution algorithms have been proposed that do not use explicitly crossover and mutation of traditional genetic algorithms, but estimate the distribution of population for more efficient search. But because it is not easy to discover higher-order correlations of variables, lower-order correlations are estimated most cases under various constraints. In this paper, we propose a new estimation of distribution algorithm that represents higher-order correlations of the data and finds global optimum more efficiently. The proposed algorithm represents the higher-order correlations among variables by building random hypergraph model composed of hyperedges consisting of variables which are expected to be correlated, and generates the next population by Bayesian sampling algorithm Experimental results show that the proposed algorithm can find global optimum and outperforms the simple genetic algorithm and BOA(Bayesian Optimization Algorithm) on decomposable functions with deceptive building blocks.

• Electronics and Telecommunications Trends
• /
• v.34 no.4
• /
• pp.32-42
• /
• 2019

The performance of machine learning algorithms significantly depends on how a configuration of hyperparameters is identified and how a neural network architecture is designed. However, this requires expert knowledge of relevant task domains and a prohibitive computation time. To optimize these two processes using minimal effort, many studies have investigated automated machine learning in recent years. This paper reviews the conventional random, grid, and Bayesian methods for hyperparameter optimization (HPO) and addresses its recent approaches, which speeds up the identification of the best set of hyperparameters. We further investigate existing neural architecture search (NAS) techniques based on evolutionary algorithms, reinforcement learning, and gradient derivatives and analyze their theoretical characteristics and performance results. Moreover, future research directions and challenges in HPO and NAS are described.

• Journal of KIISE:Software and Applications
• /
• v.33 no.8
• /
• pp.681-691
• /
• 2006

Clustering is one of widely used methods for grouping thousands of genes by their similarities of expression levels, so that it helps to analyze gene expression profiles. This method has been used for identifying the functions of genes. Fuzzy clustering method, which is one category of clustering, assigns one sample to multiple groups according to their degrees of membership. This method is more appropriate for analyzing gene expression profiles because single gene might involve multiple genetic functions. Clustering methods, however, have the problems that they are sensitive to initialization and can be trapped into local optima. To solve these problems, this paper proposes an evolutionary fuzzy clustering method, where adaptive a-cut based evaluation is used for the fitness evaluation to apply different criteria considering the characteristics of datasets to overcome the limitation of Bayesian validation method that applies the same criterion to all datasets. We have conducted experiments with SRBCT and yeast cell-cycle datasets and analyzed the results to confirm the usefulness of the proposed method.