Development of Efficient Software for Sequential Minimal Optimization

Ripepi, Giuseppe (2013) Development of Efficient Software for Sequential Minimal Optimization. Technical Report. IMATI, Genova.

[img] PDF
Report_IMATI_Ge_11_2013.pdf - Published Version
Restricted to Repository staff only

Download (1MB) | Request a copy


Sequential Minimal Optimization (SMO) is one of the most popular and fast algorithm that solves the Support Vector Machine (SVM) training problem. However, SMO requires a large amount of computation time for solving large size problems, thus the interest in developing an efficient parallel implementation. In this paper, we analyzed two different parallel implementations of this algorithm, and also two different development tools: gcc+MPICH and Intel’s icc+MPI. We experimentally evaluated that there is not an optimal solution for all the cases, in particular considering the present heterogeneous computing platform.

Item Type: Monograph (Technical Report)
Subjects: 000 Scienza degli elaboratori - Scienze dell’informazione - Scienze archivistiche, librarie e dell'informazione documentaria – opere generali
Depositing User: Ms. Cristiana Costalunga
Date Deposited: 09 Jun 2014 10:13
Last Modified: 09 Jun 2014 10:13

Actions (login required)

View Item View Item