Evgueni N. Smirnov

Learn More
We address the problem of applying machine-learning classi-fiers in domains where incorrect classifications have severe consequences. In these domains we propose to apply classifiers only when their performance can be defined by the domain expert prior to classification. The classifiers so obtained are called reliable classifiers. In the article we present(More)
Instance retraction is a difficult problem for concept learning by version spaces. In this paper, two new version-space representations are introduced: instance-based maximal boundary sets and instance-based minimal boundary sets. They are correct representations for the class of admissible concept languages and are efficiently computable. Compared to other(More)
Reliable classifiers abstain from uncertain instance classifications. In this paper we extend our previous approach to construct reliable classifiers which is based on isometrics in Receiver Operator Characteristic (ROC) space. We analyze the conditions to obtain a reliable classifier with higher performance than previously possible. Our results show that(More)
The authors propose a meta-typicalness approach to apply the typicalness framework for any type of classifiers. The approach can be used to construct classifiers with specified classification performance. Experiments show that the approach results in classifiers that can outperform an existing typicalness-based classifier