Fuzzy sets in pattern recognition: Methodology and methods

摘要

An objective of the paper is to discuss a state-of-the-art of methodology and algorithms of fuzzy sets in the field of pattern recognition. In real-world recognition and classification problems we are faced with fuzziness that is connected with diverse facets of cognitive activity of the human being. An origin of sources of fuzziness is related to labels expressed in feature space as well as to labels of classes taken into account in classification procedures. An evident difference between a way of information processing by means of probability and fuzzy set theory and a way of interpretation of results is explained in detail. In sequel methods of pattern recognition are studied in two main streams, namely supervised and unsupervised learning. Different approaches to designing of classification schemes (as e.g. relation calculus, decision-making approach, etc.) are put into account. A method of feature selection with the aid of fuzzy measure and fuzzy integral is introduced. In clustering techniques methods relying on minimization of objective function are analyzed in detail. Also a problem of cluster validity expressed in terms of clustering indices is addressed. Numerical examples are also provided.