Instance retraction is a difficult problem for concept learning by version spaces. This chapter introduces a family of version-space representations called one-sided instance-based boundary sets. They are correct and efficiently computable representations for admissible concept languages. Compared to other representations, they are the most efficient useful(1) version-space representations for instance retraction.
|Title of host publication||Database Support for Data Mining Applications|
|Subtitle of host publication||Discovering Knowledge with Inductive Queries|
|Editors||R Meo, PL Lanzi, M Klemettinen|
|Number of pages||19|
|Publication status||Published - 2004|
|Series||Lecture Notes in Computer Science|