General framework for binary nonlinear classification on top samples

0519654 - ÚTIA 2020 RIV IT eng A - Abstract
Mácha, Václav - Adam, Lukáš - Šmídl, Václav
General framework for binary nonlinear classification on top samples.
Book of Abstracts of the 3rd International Conference and Summer School, Numerical Computations: Theory and Algorithms. Rende: Centro Editoriale e Librario dell’Universit`a della Calabria, 2019 - (Sergeyev, Y.; Kvasov, D.; Mukhametzhanov, M.; Nasso, M.). s. 206-206. ISBN 9788874581016.
[NUMERICAL COMPUTATIONS: THEORY AND ALGORITHMS (NUMTA–2019). 15.06.2019-21.06.2019, Le Castella Village]
R&D Projects: GA ČR GA18-21409S
Institutional support: RVO:67985556
Keywords : binary classification * duality * kernels * accuracy at the top * ranking * hypothesis testing
OECD category: Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
http://library.utia.cas.cz/separaty/2019/AS/macha-0519654.pdf

In our previous work [1], we have proposed a general framework to handle binary linear classification for top samples. Our framework includes ranking problems, accuracy at the top or hypothesis testing. We have summarized known methods, such as [2, 3, 4], belonging to this framework and proposed new ones. Note that these methods were either derived in their primal form, or they did
not use kernels. This forced a restriction on only linear classifiers.
Permanent Link: http://hdl.handle.net/11104/0304787