楕円領域を持つファジィクラシファイアの最適データ分割による汎化能力の向上方式 Improvement of Generalization Ability of a Fuzzy Classifier with Ellipsoidal Regions by the Opeimum Division of Data
In this paper, we discuss methods for improving the generalization ability of a fuzzy classifier with ellipsoidal regions. In the fuzzy classifier, each cluster is approximated by a center and a covariance matrix, and the membership function is calculated using the inverse of the covariance matrix. Thus when the number of training data is small, the covariance matrix becomes singular and the generalization ability decreases. In addition, when the characteristics of the training and test data differ, the generalization ability decreases. In this paper, we improve the generalization ability controlling the number of singular values in the covariance matrix. Then we propose to divide the sampled data set into training and test data sets so that the centers and the covariance matrices of each class become similar. Finally, we demonstrate the validity of our methods by computer simulations.
システム制御情報学会論文誌 13(2), 87-94, 2000-02-15