Krzysztof Krawiec


Home

Research:

edit SideBar

A classification task is a test-based problem, with examples corresponding to tests. A correct classification is equivalent to passing a test, while incorrect to failing it. This applies also to classifying pixels in an image, viz. image segmentation. A natural performance indicator in such a setting is the accuracy of classification, i.e., the fraction of passed tests. When solving a classification tasks with genetic programming, it is thus common to employ this indicator as a fitness function. However, recent developments in GP as well as some earlier work suggest that the quality of evolved solutions may benefit from using other search drivers to guide the traversal of the space of programs. In this study, we systematically verify the usefulness of selected alternative search drivers in the problem of detection of blood vessels in ophthalmology imaging.

@INPROCEEDINGS { Krawiec:2015:evoApplications,
    ABSTRACT = { A classification task is a test-based problem, with examples corresponding to tests. A correct classification is equivalent to passing a test, while incorrect to failing it. This applies also to classifying pixels in an image, viz. image segmentation. A natural performance indicator in such a setting is the accuracy of classification, i.e., the fraction of passed tests. When solving a classification tasks with genetic programming, it is thus common to employ this indicator as a fitness function. However, recent developments in GP as well as some earlier work suggest that the quality of evolved solutions may benefit from using other search drivers to guide the traversal of the space of programs. In this study, we systematically verify the usefulness of selected alternative search drivers in the problem of detection of blood vessels in ophthalmology imaging. },
    ADDRESS = { Copenhagen },
    AUTHOR = { Krzysztof Krawiec and Miko{\l}aj Pawlak },
    BOOKTITLE = { 18th European Conference on the Applications of Evolutionary Computation },
    DOI = { doi:10.1007/978-3-319-16549-3_45 },
    EDITOR = { Antonio M. Mora and Giovanni Squillero },
    ISBN13 = { 978-3-319-16548-6 },
    KEYWORDS = { genetic algorithms, genetic programming, Search drivers, Binary classification, Image segmentation },
    MONTH = { {8-10 } # apr },
    NOTES = { EvoIASP EvoApplications2015 held in conjunction with EuroGP'2015, EvoCOP2015 and EvoMusArt2015 http://www.evostar.org/2015/cfp_evoapps.php },
    ORGANISATION = { EvoStar },
    PAGES = { 554--566 },
    PUBLISHER = { Springer },
    SERIES = { LNCS },
    TITLE = { Genetic Programming with Alternative Search Drivers for Detection of Retinal Blood Vessels },
    VOLUME = { 9028 },
    YEAR = { 2015 },
    1 = { https://doi.org/10.1007/978-3-319-16549-3_45 },
}


Powered by PmWiki