http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
A Design of Granular Model Using Conditional Clustering and Density Peaks
Keun-Chang Kwak(곽근창) 한국정보기술학회 2015 한국정보기술학회논문지 Vol.13 No.7
In this paper, an improved conditional clustering method based on rapid search of density peaks to construct granular model is proposed. This clustering generates linguistic contexts in the output space and estimates the cluster centers so that possess the homogeneity between input and output space. Furthermore, it can obtain the valid number of cluster corresponding to each context by rapidly searching density peaks using the correlation between local density and distance index from high density points. Thus, this proposed clustering approach can be used as if-then rules with unique characteristics of granular model. It also has the prediction performance by the model output with uncertainty. For this, the experiments are performed on simple synthetical data sets and real-world application problem to demonstrate the superiority and effectiveness through the efficient rule extraction and design of granular model.
곽근창(Keun-Chang Kwak) 한국정보기술학회 2009 한국정보기술학회논문지 Vol.7 No.1
This paper is concerned with three-dimensional object recognition by TSA (Tensor Subspace Analysis). The traditional well-known PCA (Principal Component Analysis) and LDA (Linear Discriminant Analysis) frequently used in conjunction with face and object recognition consider as a high dimensional vector represented by image, while TSA considers an image as the second order tensor. The relationship between the column and row vectors of the image matrix can be naturally characterized by TSA. Furthermore, TSA finds the intrinsic geometrical structure of the tensor space by learning a lower dimensional tensor subspace. The experimental results on COIL(Columbia object image library) object database used in this paper demonstrate that TSA showed a good recognition rate in comparison with that of PCA and LDA.
유전알고리즘과 FCM 기반 퍼지 시스템을 이용한 비선형 시스템 모델링
곽근창(Keun-chang Kwak),이대종(Dae-jong Lee),유정웅(Jeong-woong Ryu),전명근(Myung-geun Chun) 한국지능시스템학회 2001 한국지능시스템학회논문지 Vol.11 No.6
본 논문에서는 유전알고리즘(Genetic Algorithm)과 FCM(Fuzzy c-means) 클러스터링을 이용하여 TSK(Takagi-Sugeno-Kang)형태의 퍼지 규칙 생성과 퍼지 시스템(FCM-ANFIS)을 효과적으로 구축하는 방법을 제안한다. 구조동정에서는 먼저 PCA(Principal Component Analysis)을 이용하여 입력 데이터 성분간의 상관관계를 제거한 후에 FCM을 이용하여 클러스터를 생성하고 성능지표에 근거해서 타당한 클러스터의 수, 즉 퍼지 규칙의 수를 얻는다. 파라미터 동정에서는 유전알고리즘을 이용하여 전제부 파라미터를 최적에 가깝도록 탐색을 시도한다. 결론부 파라미터는 유전알고리즘에 의한 탐색공간을 줄이기 위해 전제부 파라미터가 결정되면 RLSE(Recursive Least Square Estimate)에 의해 추정되어진다. 이렇게 함으로서 타당한 규칙 수와 효율적인 퍼지 규칙을 얻을 수 있다. 제안된 방법의 유용성을 보이기 위해 Box-Jenkins의 가스로 데이터와 Rice taste 데이터의 모델링에 적용하여 이전의 연구보다 좋은 결과를 보임을 알 수 있었다. In this paper, the scheme of an efficient fuzzy rule generation and fuzzy system construction using GA(genetic algorithm) and FCM(Fuzzy c-means) clustering algorithm is proposed for TSK(Takagi-Sugeno-Kang) type fuzzy system. In the structure identification, input data is transformed by PCA(Principal Component Analysis) to reduce the correlation among input data components. And then, a set of fuzzy rules are generated for a given criterion by FCM clustering algorithm. In the parameter identification, premise parameters are optimally searched by GA. On the other hand, the consequent parameters are estimated by RLSE(Recursive Least Square Estimate) to reduce the search space. From this, one can systematically obtain the valid number of fuzzy rules which shows satisfying performance for the given problem. Finally, we applied the proposed method to the Box-Jenkins data and rice taste data modeling problems and obtained a better performance than previous works.