Stability Issues with Classifier Using Lukasiewicz Similarity and Modified Schweizer & Sklar Equations
Pasi Luukka, and Jouni Sampo
Laboratory of Applied Mathematics, Lappeenranta University of Technology P.O. Box 20, FIN-53851 Lappeenranta, Finland
In this article we have tested the stability of a classifier based on Lukasiewicz similarity in the generalized Lukasiewicz structure. We have also tested Schweizer & Sklar’s implications with an extension to generalized mean to classification task. We will show that classification results are not so sensitive to p values with Schweizer & Sklar’s measures, which indicates a generalized form of equations. In this article we have also tested the stability of these measures. Two different tests for stability were made: In one test the stability was checked with respect to weight parameters and the other test was carried out for idealvectors. The tests were done with five different classification problems.