• Communications for Statistical Applications and Methods
• /
• v.8 no.1
• /
• pp.193-207
• /
• 2001

When designing a test set, we need to consider constraints on items that are deemed important by item developers or test specialists. The constraints are essentially on the components of the test domain or abilities relevant to a given test set. And so if the test domain could be represented in a more refined form, test construction would be made in a more efficient way. We assume that relationships among task abilities are representable by a causal model and that the item response theory (IRT) is not fully available for them. In such a case we can not apply traditional item selection methods that are based on the IRT. In this paper, we use entropy as an uncertainty measure for making inferences on task abilities and developed an optimal item selection algorithm which reduces most the entropy of task abilities when items are selected from an item pool.

• Journal of KIISE:Computer Systems and Theory
• /
• v.35 no.8
• /
• pp.372-377
• /
• 2008

Multiclass cancer classification has been actively investigated based on gene expression profiles, where it determines the type of cancer by analyzing the large amount of gene expression data collected by the DNA microarray technology. Since gene expression data include many genes not related to a target cancer, it is required to select informative genes in order to obtain highly accurate classification. Conventional rank-based gene selection methods often use ideal marker genes basically devised for binary classification, so it is difficult to directly apply them to multiclass classification. In this paper, we propose a novel method for multiclass gene selection, which does not use ideal marker genes but directly analyzes the distribution of gene expression. It measures the class-discriminability by discretizing gene expression levels into several regions and analyzing the frequency of training samples for each region, and then classifies samples by using the naive Bayes classifier. We have demonstrated the usefulness of the proposed method for various representative benchmark datasets of multiclass cancer classification.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2016.07a
• /
• pp.95-96
• /
• 2016

본 논문에서는 디리클레 분포와 베이즈 추론 모델을 활용하여 전자우편을 분류하고 정리하는 방법을 제안한다. 과거 원치 않는 광고성 이메일인 스팸 탐지에서 시작한 전자우편 분류는 지속적인 송수신 량의 증가와 내용의 다양화로 인해 광고성과 정보성의 판단 기준이 모호해진 상태이다. 스팸 탐지와 같은 이분법적 분류 방식이 아닌 내용의 주제 별로 자동 분류할 수 있는 방법이 필요하다. 본 논문에서 다루는 제안 기법은 전자우편의 내용에서 다뤄질 수 있는 주제의 종류를 예측하기 위한 방법을 제공한다. 발신하거나 수신된 전자우편이 속한 주제를 자동으로 정할 수 있다. 본 제안 기법의 활용을 통해 전자우편의 분류만이 아닌 업무 및 시장 동향 분석과 정보보안 분야에서는 악성코드 분류에 사용될 수 있을 것으로 기대된다.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2005.11a
• /
• pp.205-208
• /
• 2005

We have fuzzy hypotheses testing from Bayesian statistics with ideas from fuzzy sets theory to generalize Bayesian methods both for samples of fuzzy data and for prior distributions with non-precise parameters. Appling the principle of agreement index, the posterior odds ratio in the favor of hypotheses $H_0$ is equal to product of the fuzzy odds ratio and the fuzzy likelihood ratio. If the Posterior odds ratio exceeds the grade judgement, we accept the hypothesis $H_0$ for the degree.

• Journal of Korea Multimedia Society
• /
• v.8 no.6
• /
• pp.783-789
• /
• 2005

This paper is proposed a watermarking technique for copyright protection of multimedia contents. We proposed adaptive watermark detection algorithm using stochastic perceptual model and statistical decision method in DMWT(discrete multi wavelet transform) domain. The stochastic perceptual model calculates NVF(noise visibility function) based on statistical characteristic in the DMWT. Watermark detection algorithm used the likelihood ratio depend on Bayes' decision theory by reliable detection measure and Neyman-Pearson criterion. To reduce visual artifact of image, in this paper, adaptively decide the embedding number of watermark based on DMWT, and then the watermark embedding strength differently at edge and texture region and flat region embedded when watermark embedding minimize distortion of image. In experiment results, the proposed statistical decision method based on multiwavelet domain could decide watermark detection.

• Proceedings of the KIEE Conference
• /
• 1988.11a
• /
• pp.469-472
• /
• 1988

This paper deals with signal processing and pattern recognition for the pulsed doppler radar. In order to identify the class of moving targets detected by radar, linear predictive analysis is utilized to extract reflection coefficients of each radar signal as features, and Bayes decision theory is applied to classify them.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.24 no.3
• /
• pp.504-509
• /
• 1987

In this paper a characteristics of image which can be segmented based on the thresholding technique using a histogram was investigated employing 3 parameters: the variance of pixel value, the average mean difference between target and background and the target size. The threshold value for the histogram segmentation was determined by applying the hypothesis testing theory. The performance of the selected threshold was evaluated by computing a probability of error. Since a priori probability can be easily obtained from the histogram, it was found that the Bayes decision rule which theoretically guarantees the minimum probability of error works better than the minimax criterion rule.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.12
• /
• pp.108-115
• /
• 1994

Most pattern classifiers have been designed based on the ML (Maximum Likelihood) training algorithm which is simple and relatively powerful. The ML training is an efficient algorithm to individually estimate the model parameters of each class under the assumption that all class models in a classifier are statistically independent. That assumption, however, is not valid in many real situations, which degrades the performance of the classifier. In this paper, we propose a minimum-error-rate training algorithm based on the MAP (Maximum a Posteriori) approach. The algorithm regards the normalized outputs of the classifier as estimates of the a posteriori probability, and tries to maximize those estimates. According to Bayes decision theory, the proposed algorithm satisfies the condition of minimum-error-rate classificatin. We apply this algorithm to NPM (Neural Prediction Model) for speech recognition, and derive new disrminative training algorithms. Experimental results on ten Korean digits recognition have shown the reduction of 37.5% of the number of recognition errors.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.29B no.9
• /
• pp.37-49
• /
• 1992

The standard definition of computational vision is a set of inverse problems of recovering surfaces from images. Thus the common characteristics of the most early vision problems are ill-posed. The main idea for solving ill-posed problems is to restrict the class of admissible solutions by introducing suitable a priori knowledge. Standard regurarization methods lead to satisfactory solutions of early vision problems but cannot deal effectively and directly with a few general problems, such as discontinuity and fusion of information from multiple modules. In this paper, we discuss limitations of standard regularization theory and present new stochastic method. We will outline a rigorous approach to overcome part of ill-posedness of image restoration, edge detection, and stereo vision problems, based on Bayes estimation and MRF(Markov random field) model, that effectively deals with the problems. This result makes one hope that this framework could be useful in the solution of other vision problems.

• Journal of the Korean Society of Safety
• /
• v.10 no.2
• /
• pp.10-16
• /
• 1995

In estimating the fatigue data, only the uniform safety rate has been applied so far However, since more reasonable design concepts such as machine structures or subsidiary materials will be required in the future, the importance of a statistical estimation method for fatigue data is being highlighted. With such basic conception in mind, this study was aimed at critically discussing the interval estimation method which has been applied using the classical statistics thus far It was conceived that this conventional method would result in the estimation of the unstable side from the viewpoint of the likelihood Interval estimation method. In this regard, this study aimed at estimating the fatigue strength through the likelihood interval estimation method comparing it with the conventional interval estimation method would result in the estimation of the unstable side from the viewpoint of the likelihood interval estimation method. One of the methods using the likelihood for estimation data is the Bayes method. Based on this theory, statistical estimations were positivly applied, and thereupon, the fatigue data were estimated.