A. Bánhalmi (Hungary)
One-class classification, artificial counter-example genera tion.
Here we propose novel methods for the One-Class Clas sification task and examine their applicability. Essentially, these methods extend the training set – which contains only positive examples – with artificially generated counter examples. After, a two-class classifier is used to separate them. In this paper following a description of the exist ing and the newly proposed methods some problematic is sues are investigated theoretically and studied empirically by applying these methods to artificial datasets. Then their efficiency is compared to those of other one-class classifi cation methods.
Important Links:
Go Back
