Monotone Classification by Function Decomposition



Function decomposition in the context of classification focuses on splitting a concept (represented by the data set) into less complex sub-concepts thus revealing implicit information about the underlying structure of the problem. The sub-concepts can then be used to generate classification rules. In the case of monotone problems a requirement might be posed for the resulting classifier to be monotone. This talk presents a method for generating a monotone decomposition and a monotone classifier in an automated way without using any additional expert knowledge, predefined operators or intermediate functions. Unlike the decomposition methods for the general case, here a default rule is defined which ensures that the whole input space is covered.
Download sheets
Contact information:
Wilco van den Heuvel