PhD, Senior Lecturer at Institute of Computing Science at the Poznan University of Technology.

My research interests include machine learning, especially cost-efficient feature selection. The costs of acquiring feature values of classified objects (test costs) are considered. These costs can be reduced by feature selection. Standard approaches to feature selection select a common subset of features for all classified objects (a global approach). An alternative approach is to select a feature set for each classified object separately (a local approach). Such a local approach is sequential classification which is an extension of its base classifier by a feature selection strategy and a stopping criterion. A proper selection of these two elements allows one to compete with global approaches. The use of levels of abstractions for attribute values in order to further reduce these test costs is proposed.

See my profile at Google Scholar