Investigating convergence of linear SVM implemented in PermonSVM employing MPRGP algorithm

0495870 - ÚGN 2019 RIV CH eng C - Konferenční příspěvek (zahraniční konf.)
Kružík, Jakub - Pecha, Marek - Hapla, D. - Horák, David - Čermák, Martin
Investigating convergence of linear SVM implemented in PermonSVM employing MPRGP algorithm.
High Performance Computing in Science and Engineering. HPCSE 2017. Cham: Springer, 2018 - (Kozubek, T.), s. 115-129. Lecture Notes in Computer Science, Code 216349, Volume 11087. ISBN 978-3-319-97135-3.
[HPCSE 2017: International Conference on High Performance Computing in Science and Engineering /3./. Karolinka (CZ), 22.05.2017-25.05.2017]
Grant CEP: GA MŠMT LQ1602
Grant ostatní: Ga MŠk(CZ) LM2015070; GA ČR(CZ) GA15-18274S
Institucionální podpora: RVO:68145535
Klíčová slova: MPRGP * PERMON * PermonQP * PermonSVM * quadratic programming * support vector machines
Obor OECD: Applied mathematics
https://link.springer.com/chapter/10.1007/978-3-319-97136-0_9

This paper deals with the novel PermonSVM machine learning tool. PermonSVM is a part of our PERMON toolbox. It implements the linear two-class Support Vector Machines. PermonSVM is built on top of PermonQP (PERMON module for quadratic programming) which in turn uses PETSc. The main advantage of PermonSVM is that it is parallel. The parallelism comes from a distribution of matrices and vectors. The MPRGP algorithm, implemented in PermonQP, is used as a solver of the quadratic programming problem arising from the dual SVM formulation. The scalability of MPRGP was proven in problems of mechanics with more than billion of unknowns solved on tens of thousands of cores. Apart from the scalability of our approach, we also investigate the relations between training rate, hyperplane margin, the value of the dual functional, and the norm of the projected gradient.
Trvalý link: http://hdl.handle.net/11104/0288753