Number of the records: 1
Investigating convergence of linear SVM implemented in PermonSVM employing MPRGP algorithm
- 1.
SYSNO ASEP 0495870 Document Type C - Proceedings Paper (int. conf.) R&D Document Type Conference Paper Title Investigating convergence of linear SVM implemented in PermonSVM employing MPRGP algorithm Author(s) Kružík, Jakub (UGN-S)
Pecha, Marek (UGN-S)
Hapla, D. (CZ)
Horák, David (UGN-S) SAI, ORCID
Čermák, Martin (UGN-S)Number of authors 5 Source Title High Performance Computing in Science and Engineering. HPCSE 2017. - Cham : Springer, 2018 / Kozubek T. - ISBN 978-3-319-97135-3 Pages s. 115-129 Number of pages 15 s. Publication form Online - E Action HPCSE 2017: International Conference on High Performance Computing in Science and Engineering /3./ Event date 22.05.2017 - 25.05.2017 VEvent location Karolinka Country CZ - Czech Republic Event type WRD Language eng - English Country CH - Switzerland Keywords MPRGP ; PERMON ; PermonQP ; PermonSVM ; quadratic programming ; support vector machines Subject RIV BA - General Mathematics OECD category Applied mathematics R&D Projects LQ1602 GA MŠMT - Ministry of Education, Youth and Sports (MEYS) Institutional support UGN-S - RVO:68145535 UT WOS 000469334300009 EID SCOPUS 85050502411 DOI 10.1007/978-3-319-97136-0_9 Annotation 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. Workplace Institute of Geonics Contact Lucie Gurková, lucie.gurkova@ugn.cas.cz, Tel.: 596 979 354 Year of Publishing 2019 Electronic address https://link.springer.com/chapter/10.1007/978-3-319-97136-0_9
Number of the records: 1