• 한국음향학회지
• /
• 제26권5호
• /
• pp.194-198
• /
• 2007

본 논문에서는 음성의 통계적 모델에 기반한 음성검출기의 성능향상을 위해 변별적 가중치 학습(discriminative weight training) 기반의 최적화된 우도비 테스트(Likelihood Ratio Test, LRT)를 제안한다. 먼저, 기존의 통계모델기반의 음성검출기를 분석하고, 이를 기반으로 MCE(minimum classification error)방법을 도입하여, 각 주파수 채널별로 다른 가중치를 가지는 우도비 기반의 음성검출 결정법(decision rule)을 제시한다. 제안된 알고리즘은 비정상(non-stationary)잡음환경에서 기존의 동일 가중치를 가지는 기하 평균 기반의 음성검출기와 비교하였으며, 우수한 성능을 보인다.

• 조명전기설비학회논문지
• /
• 제20권9호
• /
• pp.45-53
• /
• 2006

This paper proposes a method for an effective intelligent route decision for automatic moving robots(AMR) using a 2D spatial map of a stereo camera system. In this method, information about depth and disparity map are detected in the inputting images of a parallel stereo camera. The distance between the automatic moving robot and the obstacle is detected, and a 2D spatial map is obtained from the location coordinates. Then the relative distances between the obstacle and other objects are deduced. The robot move automatically by effective and intelligent route decision using the obtained 2D spatial map. From experiments on robot driving with 240 frames of stereo images, it was found that the error ratio of the calculated distance to the measured distance between objects was very low, 1.52[%] on average.

• 한국경영과학회:학술대회논문집
• /
• 대한산업공학회/한국경영과학회 1992년도 춘계공동학술대회 발표논문 및 초록집; 울산대학교, 울산; 01월 02일 May 1992
• /
• pp.117-126
• /
• 1992

This study proposes a new interactive multicriteria method for determining the best levels of the decision variables needed to optimize a stochastic computer simulation with multiple response variables. The method, called the Pairwise Comparison Stochastic Cutting Plane (PCSCP) method, combines good features from interactive multiple objective mathematical programming methods and response surface methodology. The major characteristics of the PCSCP algorithm are: (1) it interacts progressively with the decision maker (DM) to obtain his preferences, (2) it uses good experimental design to adequately explore the decision space while reducing the burden on the DM, and (3) it uses the preference information provided by the DM and the sampling error in the responses to reduce the decision space. This paper presents the basic concepts of the PCSCP method along with its performance for solving randomly selected test problems.

• 한국정밀공학회:학술대회논문집
• /
• 한국정밀공학회 2006년도 춘계학술대회 논문집
• /
• pp.615-616
• /
• 2006

Many manufacturing companies manage their production-related data for quality management and production management. Nevertheless, production related-data should be closely related to each other Stored data is mainly used to monitor their process and products' error. In this paper, we provide an enterprise-wide production data model for decision support system and product automation. Process data, quality-related data, and test data are integrated to identify the process inter or intra dependency, the yield forecasting, and the trend of process status. In addition, it helps the manufacturing decision support system to decide critical manufacturing problems.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• 제3권1호
• /
• pp.44-51
• /
• 2003

Knowledge acquisition is a bottleneck in knowledge-based system implementation. Decision tree induction is a useful machine learning approach for extracting classification knowledge from a set of training examples. Many real-world data contain fuzziness due to observation error, uncertainty, subjective judgement, and so on. To cope with this problem of real-world data, there have been some works on fuzzy classification rule learning. This paper makes a survey for the kinds of fuzzy classification rules. In addition, it presents a fuzzy classification rule learning method based on decision tree induction, and shows some experiment results for the method.

• 정보학연구
• /
• 제2권1호
• /
• pp.103-113
• /
• 1999

본 논문에서는 연판정지향 알고리즘에 Stop-and-Go 알고리즘의 개념을 결합한 연판정지향 Stop-and-Go 알고리즘을 제안한다. 제안한 알고리즘은 두 알고리즘보다 더 신뢰성 있는 오차신호를 사용함으로써 향상된 등화 성능을 가진다. 컴퓨터 모의실험을 통하여 제안한 알고리즘이 CMA, MCMA, Stop-and-Go 알고리즘, 단순화된 연판정지향 알고리즘에 비해 잔류 심벌간 간섭과 정상상태로의 수렴 속도면에서 우수한 성능을 가짐을 확인하였다.

• 한국산학기술학회논문지
• /
• 제22권4호
• /
• pp.580-586
• /
• 2021

현재 사출성형분야의 Computer Aided Testing(CAT) 방법론으로 CAE(Computer Aided Engineering)를 이용한 수치 해석 기법이 주를 이루고 있다. 그러나 최근 시뮬레이션에 추가로 인공지능 기법을 응용하는 방법론이 연구되고 있다. 우리는 지난 연구에서 다양한 Machine Learning 기법을 활용하여 사출 성형 공정에 따른 변형 결과를 비교하였으며, 최종적으로 MLP(Multi-Layer Perceptron) 예측모델을 생성하였고, HMA(Hybrid Metaheuristic Algorithm)를 이용하여 최적화 결과를 얻어냈다. 그러나 MLP는 예측 성능이 우수한 반면 블랙박스와 같이 결정 과정에 대한 설명이 부족하다. 본 연구에서는 Radiator Tank 부품에 대하여 사출 성형 해석 소프트웨어인 Autodesk Moldflow 2018을 이용하여 수치 해석 기법으로 데이터를 생성하고, Machine Learning 소프트웨어인 RapidMiner Studio version 9.5를 활용하여 여러 Machine Learning Algorithms 모델을 생성하여 평균 제곱근 오차를 비교하였다. Decision-tree는 Root Mean Square Error(RMSE) 값이 다른 Machine Learning 기법에 비해 양호한 예측 성능을 갖추고 있었다. Decision-tree의 크기를 결정하는 Maximal Depth에 따라 분류 기준을 높일 수 있지만 복잡성도 함께 증가시켰다. Decision-tree를 이용하여 구속 조건을 만족하는 중간 값을 선정하여 시뮬레이션을 진행한 결과 기존의 시뮬레이션만 진행한 것보다 7.7%의 개선 효과가 있었다.

• Journal of the Korean Data and Information Science Society
• /
• 제17권2호
• /
• pp.269-278
• /
• 2006

This paper presents the new hybrid data mining technique using error pattern, modeling of improving classification accuracy. The proposed method improves classification accuracy by combining two different supervised learning methods. The main algorithm generates error pattern modeling between the two supervised learning methods(ex: Neural Networks, Decision Tree, Logistic Regression and so on.) The Proposed modeling method has been applied to the simulation of 10,000 data sets generated by Normal and exponential random distribution. The simulation results show that the performance of proposed method is superior to the existing methods like Logistic regression and Discriminant analysis.

• 한국경영과학회지
• /
• 제30권4호
• /
• pp.27-43
• /
• 2005

This paper presents a new hybrid data mining technique using error pattern modeling to improve classification accuracy when the data type of a target variable is binary. The proposed method increases prediction accuracy by combining two different supervised learning methods. That is, the algorithm extracts a subset of training cases that are predicted inconsistently by both methods, and models error patterns from the cases. Based on the error pattern model, the Predictions of two different methods are merged to generate final prediction. The proposed method has been tested using practical 10 data sets. The analysis results show that the performance of proposed method is superior to the existing methods such as artificial neural networks and decision tree induction.

• 대한전자공학회논문지
• /
• 제26권11호
• /
• pp.1819-1828
• /
• 1989

In this paper, we determined the modulation index for minimization of speed ripple and position error using maximum speed ripple (SRF) and maximum position error(PEF) in current source TPWM inverter. Through computer simulation, we compared with total current harmonic distortion, SRF and PEF for square wave modulation method and TPWM method. As a result, it turns out that square wave modulation method is superior to TPWM method of 3 pulses per half cycle in speed ripple and position error contents. And TPWM is better than square wave method when pulse number is more than 5. Also, in these pulse numbers, moduladtion index of minimum speed ripple and munimum position error is 0.91.