• Proceedings of the Korean Information Science Society Conference
• /
• 2005.07b
• /
• pp.628-630
• /
• 2005

베이지안 네트워크는 확률이론에 기초해 불확실성이 존재하는 실세계 문제를 해결하는데 많은 기여를 하고 있다. 최근 네트워크 구조를 데이터로부터 자동으로 학습하는 많은 연구가 이루어져 보다 손쉽게 많은 사람들이 사용할 수 있게 되었다. 하지만 한번 학습하여 고정된 네트워크의 구조는 새롭게 수집되는 데이터의 특성을 잘 반영하지 못하는 문제를 지니고 있다. 환경의 변화에 맞게 지속적으로 네트워크 구조를 갱신하기 위한 연구가 진행되고 있으며 본 연구에서는 Lam이 제안한 MDL기반 평가함수를 이용한 진화적 갱신 방법을 제안하여 갱신 성능을 향상시키고자 한다. 벤치마크 네트워크인 ASIA에 대한 실험 결과 제안한 방법이 기존의 지역적 탐색 방법에 비해 향상된 성능을 제공함을 확인하였다.

• Journal of the Korea Society of Computer and Information
• /
• v.14 no.11
• /
• pp.159-167
• /
• 2009

The vehicles play a significant role in modern people's life as economy grows. The development of car navigation system(CNS) provides various convenience because it shows the driver where they are and how to get to the destination from the point of source. However, the existing map-based CNS does not consider any environments such as traffic congestion. Given the same starting point and destination, the system always provides the same route and the required time. This paper proposes a path planning method with traffic prediction by applying historical driving information to the Fuzzy theory and Bayesian update. Fuzzy theory classifies the historical driving information into groups of leaving time and speed rate, and the traffic condition of each time zone is calculated by Bayesian update. An ellipse area including starting and destination points is restricted in order to reduce the calculation time. The accuracy and practicality of the proposed scheme are verified by several experiments and comparisons with real navigation.

• Korean Journal of Logic
• /
• v.14 no.3
• /
• pp.27-59
• /
• 2011

This paper concerns several versions of conditionalization. In particular, I will examine the relationship between Jeffrey conditionalization and the second order conditionalization concerning updating higher order credences. In section 2, I suggest explicitly what Jeffrey conditionalization and the second order conditionalization are. I will argue in section 3 that Jeffrey conditionalization conflicts with van Fraassen's Reflection Principle while the second order conditionalization doesn't. And I will also argue in section 4 that under some situations, Jeffrey conditionalization may lead agents to Moorean absurdity while the second conditionalization may not. As a result, I will claim that Jeffrey conditionalization is better than the second order contionalization at updating our higher order credences.

• Journal of the Korean Data and Information Science Society
• /
• v.12 no.2
• /
• pp.11-25
• /
• 2001

There are several algorithms for classification in modeling relations, patterns, and rules which exist in data. We learn to classify objects on the basis of instances presented to us, not by being given a set of classification rules. The Bayesian learning uses the probability distribution to express our knowledge about unknown parameters and update our knowledge by the law of probability as the evidence gathered from data. Also, the neural network models are designed for predicting an unknown category or quantity on the basis of known attributes by training. In this paper, we compare the misclassification error rates of Bayesian Neural Network method with those of other classification algorithms, CHAID, CART, and QUBST using several data sets.

• The Journal of the Korea Contents Association
• /
• v.20 no.6
• /
• pp.1-8
• /
• 2020

The use of text data in big data analytics has been increased. So, much research on methods for text data analysis has been performed. In this paper, we study Bayesian learning based on conjugate prior for analyzing keyword data extracted from text big data. Bayesian statistics provides learning process for updating parameters when new data is added to existing data. This is an efficient process in big data environment, because a large amount of data is created and added over time in big data platform. In order to show the performance and applicability of proposed method, we carry out a case study by analyzing the keyword data from real patent document data.

• Journal of the Korean Data and Information Science Society
• /
• v.12 no.1
• /
• pp.61-69
• /
• 2001

We present a probabilistic reasoning method for inferring knowledge about mathematical truth before an automated theorem prover completes a proof. We use a Bayesian analysis to update beleif in truth, given theorem-proving progress, and show how decision-theoretic methods can be used to determine the value of continuing to deliberate versus taking immediate action in time-critical situations.

• Journal of the Korean Geotechnical Society
• /
• v.26 no.7
• /
• pp.161-170
• /
• 2010

For the development of load and resistance factor design, reliability analysis is required to calibrate resistance factors in the framework of reliability theory. The distribution of measured-to-predicted pile resistance ratio was obrained based on only the results of load tests conducted to failure for the assessment of uncertainty regarding pile resistance and used in the conventional reliability analysis. In other words, successful pile load test (piles resisted twice their design loads without failure) results were discarded, and therefore, were not reflected in the reliability analysis. In this paper, a new systematic method based on Bayesian theory is used to update reliability indices of driven steel pipe piles by adding more proof pile load test results, even not conducted to failure, to the prior distribution of pile resistance ratio. Fifty seven static pile load tests performed to failure in Korea were compiled for the construction of prior distribution of pile resistance ratio. The empirical method proposed by Meyerhof is used to calculate the predicted pile resistance. Reliability analyses were performed using the updated distribution of pile resistance ratio. The challenge of this study is that the distribution updates of pile resistance ratio are possible using the load test results even not conducted to failure, and that Bayesian updates are most effective when limited data are available for reliability analysis.

• The KIPS Transactions:PartB
• /
• v.8B no.3
• /
• pp.231-236
• /
• 2001

일반적인 다층 신경망에서 가중치의 갱신 알고리즘으로 사용하는 오류 역전과 방식은 가중치 갱신 결과를 고정된(fixed) 한 개의 값으로 결정한다. 이는 여러 갱신의 가능성을 오직 한 개의 값으로 고정하기 때문에 다양한 가능성들을 모두 수용하지 못하는 면이 있다. 하지만 모든 가능성을 확률적 분포로 표현하는 갱신 알고리즘을 도입하면 이런 문제는 해결된다. 이러한 알고리즘을 사용한 베이지안 신경망 모형(Bayesian Neural Networks Models)은 주어진 입력값(Input)에 대해 블랙 박스(Black-Box)와같은 신경망 구조의 각 층(Layer)을 거친 출력값(Out put)을 계산한다. 이 때 주어진 입력 데이터에 대한 결과의 예측값은 사후분포(posterior distribution)의 기댓값(mean)에 의해 계산할 수 있다. 주어진 사전분포(prior distribution)와 학습데이터에 의한 우도함수(likelihood functions)에 의해 계산한 사후확률의 함수는 매우 복잡한 구조를 가짐으로 기댓값의 적분계산에 대한 어려움이 발생한다. 따라서 수치해석적인 방법보다는 확률적 추정에 의한 근사 방법인 몬테 칼로 시뮬레이션을 이용할 수 있다. 이러한 방법으로서 Hybrid Monte Carlo 알고리즘은 좋은 결과를 제공하여준다(Neal 1996). 본 논문에서는 Hybrid Monte Carlo 알고리즘을 적용한 신경망이 기존의 CHAID, CART 그리고 QUEST와 같은 여러 가지 분류 알고리즘에 비해서 우수한 결과를 제공하는 것을 나타내고 있다.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.25 no.4
• /
• pp.355-362
• /
• 2012

Remaining useful life(RUL) prediction of a system is important in the prognostics field since it is directly linked with safety and maintenance scheduling. In the physics-based prognostics, accurately estimated model parameters can predict the remaining useful life exactly. It, however, is not a simple task to estimate the model parameters because most real system have multivariate model parameters, also they are correlated each other. This paper presents representative methods to estimate model parameters in the physics-based prognostics and discusses the difference between three methods; the particle filter method(PF), the overall Bayesian method(OBM), and the sequential Bayesian method(SBM). The three methods are based on the same theoretical background, the Bayesian estimation technique, but the methods are distinguished from each other in the sampling methods or uncertainty analysis process. Therefore, a simple physical model as an easy task and the Paris model for crack growth problem are used to discuss the difference between the three methods, and the performance of each method evaluated by using established prognostics metrics is compared.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.12 no.5
• /
• pp.385-392
• /
• 2002

Collaborative filtering method for personalization can suggest new items and information which a user hasn t expected. But there are some problems. Not only the steps for calculating similarity value between each user is complex but also it doesn t reflect user s interest dynamically when a user input a query. In this paper, classifying users by their interest makes calculating similarity simple. We propose the a1gorithm for readjusting user s interest dynamically using the profile and Bayesian learning. When a user input a keyword searching for a item, his new interest is readjusted. And the user s profile that consists of used key words and the presence frequency of key words is designed and used to reflect the recent interest of users. Our methods of adjusting user s interest using the profile and Bayesian learning can improve the real satisfaction of users through the experiment with data set, collected in University s library. It recommends a user items which he would be interested in.