• Journal of the Korean Institute of Intelligent Systems
• /
• v.17 no.6
• /
• pp.799-803
• /
• 2007

In this paper, we propose a new fuzzy membership function for FSVM(Fuzzy Support Vector Machines). We apply a fuzzy membership to each input point of SVM and reformulate SVM into fuzzy SVM (FSVM) such that different input points can make different contributions to the learning of decision surface. The proposed method enhances the SVM in reducing the effect of outliers and noises in data points. This paper compares classification and estimated performance of SVM, FSVM(1), and FSVM(2) model that are getting into the spotlight in time series prediction.

• The Proceedings of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.9 no.3
• /
• pp.97-103
• /
• 1995

In this paper, for fuzzy control of linear servo system using the moving coil type linear DC motor, we propose a new fuzzy control method using parameter tuning for membership functions. A proposed fuzzy control method tunes parameters of membership function to have an appropriate control input signal for system when error exceeds predefined value and makes an inference using conventional fuzzy control rules when error reduces to a predefined value. To verify usefulness of a proposed fuzzy control method, making simulation and experiment, we compare with characteristics for conventional fuzzy control method.

• International Journal of Control, Automation, and Systems
• /
• v.2 no.3
• /
• pp.298-309
• /
• 2004

In this paper, a new evolutionary scheme to design a TSK fuzzy model from relevant data is proposed. The identification of the antecedent rule parameters is performed via the evolutionary algorithm with the unique fitness function and the various evolutionary operators, while the identification of the consequent parameters is done using the least square method. The occurrence of the multiple overlapping membership functions, which is a typical feature of unconstrained optimization, is resolved with the help of the proposed fitness function. The proposed algorithm can generate a fuzzy model with transparent membership functions. Through simulations on various problems, the proposed algorithm found a TSK fuzzy model with better accuracy than those found in previous works with transparent partition of input space.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2005.04a
• /
• pp.297-300
• /
• 2005

In this study, we proposed genetically optimized self-organizing fuzzy polynomial neural network based on information granulation and evolutionary algorithm (gdSOFPNN), develop a comprehensive design methodology involving mechanisms of genetic optimization. The proposed gdSOFPNN gives rise to a structural Iy and parametrically optimized network through an optimal parameters design available within FPN (viz. the number of input variables, the order of the polynomial, input variables, the number of membership functions, and the apexes of membership function). Here, with the aid of the information granulation, we determine the initial location (apexes) of membership functions and initial values of polynomial function being used in the premised and consequence part of the fuzzy rules respectively. The performance of the proposed gdSOFPNN is quantified through experimentation that exploits standard data already used in fuzzy modeling.

• Journal of Korean Institute of Industrial Engineers
• /
• v.23 no.2
• /
• pp.371-383
• /
• 1997

This paper propose a fuzzy regression method using fuzzy neural networks when a membership value is attached to each input-output pair. First, a method of linear fuzzy regression analysis is described by interpreting the reliability of each input-output pair as its membership values. Next, an architecture of fuzzy neural networks with fuzzy weights and fuzzy biases is shown. The fuzzy neural network maps a crisp input vector to a fuzzy output. A cost function is defined using the fuzzy output from the fuzzy neural network and the corresponding target output with a membership value. A learning algorithm is derived from the cost function. The derived learning algorithm trains the fuzzy neural network so that the level set of the fuzzy output includes the target output. Last, the proposed method is illustrated by computer simulations on numerical examples.

• Annual Conference on Human and Language Technology
• /
• 1990.11a
• /
• pp.191-198
• /
• 1990

본 논문에서는 문서 자동 인식 시스템에서 다중 카테고리로 모호하게 인식되어 질 수 있는 조합 심볼을 하나의 메타 심볼로 간주하고, 이 심볼을 fuzzy set theory에 기초를 두어 분석을 하였다. 분석 과정에서는 메타 심볼이 갖는 프리미티브들의 기울기와 길이, 프리미티브들간의 연결 및 프리미티브의 위치등의 어트리뷰트들을 이용하였다. 모호성을 내재하고 있는 메타 심볼들을 ACS(Ambiguous Category Set)의 원소로 간주하였으며, ACS의 원소들은 모호성의 원인을 제공하는 부분패턴들을 공동으로 포함하고 있다. 부분패턴을 구성하고 있는 프리미티브를 분리하여 어트리뷰트 값을 측정하고, 정의한 MF(Membership 함수)의 파라메터로 사용하였다. MF에서 얻어진 MFV(Membership Function Value)는 모호한 메타 심볼이 어떤 카테고리로 분류될 수 있는지를 나타내도록 하였다.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1993.06a
• /
• pp.1414-1417
• /
• 1993

Wang and Medel proved (1991) that fuzzy systems with product inference, centroid defuzzification, and everywhere positive membership functions (in particular, Gaussians, Wang, 1992) are capable of approximating any real continuous control function on a compact set to arbitrary accuracy. Kosko (1992) proved that fuzzy systems, in which membership functions have compact support, and combination operation (V-operation) for rules is the sum, are also universal approximators. In this paper, we generalize this result of Kosko and prove that for any &- and V-operations, any defuzzification procedure, and any basic membership function with a compact support, the resulting fuzzy controls are universal approximators. Also, Wang's result is transfered to min-inference.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2006.11a
• /
• pp.19-22
• /
• 2006

Type-2 fuzzy 이론은 기존의 퍼지 이론보다 패턴의 불확실성에 대한 제어를 더 향상시킬 수 있다. 반면에 계산 량이 커지는 문제점 때문에 본 논문에서는 type-2 fuzzy set 대신에 secondary membership이 interval의 형태를 갖는 interval type-2 fuzzy set을 기존의 radial basis function(RBF) neural network에 적용시킨 interval type-2 fuzzy RBF neural network를 제안한다. 제안한 알고리즘은 interval type-2 fuzzy membership function에 의하여 패턴들의 불확실성을 좀 더 잘 제어하여 기존의 RBF neural network의 성능을 향상시킬 수 있다. 본 논문에서는 제안한 알고리즘의 타당성을 보이기 위하여 여러 데이터 집합에 대한 분류 결과를 보인다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.4
• /
• pp.508-514
• /
• 2021

In this paper, to propose a classifier based on supervised learning, three types of fuzzy membership functions that determine the membership of each feature of classification data are proposed. In addition, the possibility of improving the classifier performance was suggested by using the average value calculation method used in the process of deriving the classification result using the average value of the membership degrees for each feature, not by using a simple arithmetic average, but by using a weighted average using various weights. To experiment with the proposed methods, three standard data sets were used: Iris, Ecoli, and Yeast. As a result of the experiment, it was confirmed that evenly excellent classification performance can be obtained for data sets of different characteristics. It was confirmed that better classification performance is possible through improvement of fuzzy membership functions and the weighted average methods.

• The Transactions of the Korean Institute of Electrical Engineers P
• /
• v.66 no.4
• /
• pp.194-199
• /
• 2017

Power transformers are an important factor for power transmission and cause fatal losses if faults occur. Various diagnostic methods have been applied to predict the failure and to identify the cause of the failure. Typical diagnostic methods include the IEC diagnostic method, the Duval diagnostic method, the Rogers diagnostic method, and the Doernenburg diagnostic method using the ratio of the main gas. However, each diagnostic method has a disadvantage in that it can't diagnose the state of the power transformer unless the gas ratio is within the defined range. In order to solve these problems, we propose a diagnosis method using ELM based intelligent algorithm and fuzzy membership function. The final diagnosis is performed by multiplying the result of diagnosis in the four diagnostic methods (IEC, Duval, Rogers, and Doernenburg) by the fuzzy membership values. To show its effectiveness, the proposed fault diagnostic system has been intensively tested with the dissolved gases acquired from various power transformers.