Search

A weighted order statistic filter (WOSF) generates a Boolean function by the specified parameters. However, WOSF with different set of parameters may generate the same Boolean function. In this paper, one proposes an alternative representation for WOS filters to characterize WOSF. This method is based on the maximal margin classification of SVM. From a truth table generated by a linearly separable Boolean function, one takes the inputs and outputs to form a training set. By SVM, it generates a maximal margin hyperplane. The hyperplane has an optimal normal vector and optimal bias, and defines a discriminant function. The discriminant function decides the category of training data and can be used to represent WOSF. In other words, one merely utilizes a normal vector and bias to represent all WOS filters with same output but different weight vectors and threshold values. [ABSTRACT FROM AUTHOR]