Inductive learning systems have been successfully applied in a number of medical domains. Nevertheless, the effective use of these systems requires data preprocessing before apply...
Mykola Pechenizkiy, Alexey Tsymbal, Seppo Puuronen
We show that the relevant information of a supervised learning problem is contained up to negligible error in a finite number of leading kernel PCA components if the kernel matche...
Mikio L. Braun, Joachim M. Buhmann, Klaus-Robert M...
Feature selection using the naive Bayes rule is presented for the case of multiclass data sets. In this paper, the EM algorithm is applied to each class projected over the feature...
The Support Vector Machine error bound is a function of the margin and radius. Standard SVM algorithms maximize the margin within a given feature space, therefore the radius is fi...
Various text mining algorithms require the process of feature selection. High-level semantically rich features, such as figurative language uses, speech errors etc., are very prom...