Similarity-based methods: a general framework for classification, approximation and association

Similarity-based methods (SBM) are a generalization of the minimal distance (MD) methods which form a basis of several machine learning and pattern re cognition methods. Investigation of similarity leads to a fruitful framewor k in which many classification, approximation and association methods are a ccommodated. Probability p(C\X; M) of assigning class C to a vector X, give n a classification model M, depends on adaptive parameters and procedures u sed in construction of the model. Systematic overview of choices available for model building is presented and numerous improvements suggested. Simila rity-Based Methods have natural neural-network type realizations. Such neur al network models as the Radial Basis Functions (RBF) and the Multilayer Pe rceptrons (MLPs) are included in this framework as special cases. SBM may a lso include several different submodels and a procedure to combine their re sults. Many new versions of similarity-based methods are derived from this framework. A search in the space of all methods belonging to the SBM framew ork finds a particular combination of parameterizations and procedures that is most appropriate for a given data. No single classification method can beat this approach. Preliminary implementation of SBM elements tested on a real-world datasets gave very good results.