摘要 : A novel approach to nonlinear classification is presented, in thetraining phase of the classifier, the training data is first clusteredin an unsupervised way by fuzzy c-means or a similar algorithm. Theclass labels are not used in... 展开 A novel approach to nonlinear classification is presented, in thetraining phase of the classifier, the training data is first clusteredin an unsupervised way by fuzzy c-means or a similar algorithm. Theclass labels are not used in this step. Then, a fuzzy relation betweenthe clusters and the class identifiers is computed. This approach allowsthe number of prototypes to be independent of the number of actualclasses. For the classification of unseen patterns, the membershipdegrees of the feature vector in the clusters are first computed byusing the distance measure of the clustering algorithm. Then, the outputfuzzy set is obtained by relational composition. This fuzzy set containsthe membership degrees of the pattern in the given classes. A crispdecision is obtained by defuzzification, which gives either a singleclass or a “reject” decision, when a unique class cannot beselected based on the available information. The principle of theproposed method is demonstrated on an artificial data set and theapplicability of the method is shown on the identification of live-stockfrom recorded sound sequences. The obtained results are compared withtwo other classifiers 收起