Robust Reductions from Ranking to Classification

Autor:	Balcan, M.F., Bansal, N., Beygelzimer, A., Coppersmith, D., Langford, J., Sorkin, G.B., Bshouty, N.H., Gentile, C.
Rok vydání:	2007
Předmět:	Core (game theory) Binary classification Receiver operating characteristic Statistics Binary number Regret Mathematics Ranking (information retrieval)
Zdroj:	Learning Theory ISBN: 9783540729259 COLT Learning theory (20th Annual Conference, COLT 2007, San Diego CA, USA, June 13–15, 2007. Proceedings), 604-619 STARTPAGE=604;ENDPAGE=619;TITLE=Learning theory (20th Annual Conference, COLT 2007, San Diego CA, USA, June 13–15, 2007. Proceedings)
DOI:	10.1007/978-3-540-72927-3_43
Popis:	We reduce ranking, as measured by the Area Under the Receiver Operating Characteristic Curve (AUC), to binary classification. The core theorem shows that a binary classification regret of r on the induced binary problem implies an AUC regret of at most 2r. This is a large improvement over approaches such as ordering according to regressed scores, which have a regret transform of $r \mapsto nr$ where n is the number of elements.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_dedup___::3d8735d51bb5adaa5a65aa0c79c36c81 https://doi.org/10.1007/978-3-540-72927-3_43 Zobrazit plný text záznamu