By Topic

Combining Meta-Learning with Multi-objective Particle Swarm Algorithms for SVM Parameter Selection: An Experimental Analysis

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

4 Author(s)
Miranda, P.B.C. ; Univ. Fed. de Pernambuco, Recife, Brazil ; Prudencio, R.B.C. ; Carvalho, A.C.P.L.F. ; Soares, C.

Support Vector Machines (SVMs) have become a well succeeded technique due to the good performance it achieves on different learning problems. However, the SVM performance depends on adjustments of its parameters' values. The automatic SVM parameter selection is treated by many authors as an optimization problem whose goal is to find a suitable configuration of parameters for a given learning problem. This work performs a comparative study of combining Meta-Learning (ML) and Multi-Objective Particle Swarm Optimization (MOPSO) techniques for the SVM parameter selection problem. In this combination, configurations of parameters provided by ML are adopted as initial search points of the MOPSO techniques. Our hypothesis is that, starting the search with reasonable solutions will speed up the process performed by the MOPSO techniques. In our work, we implemented three MOPSO techniques applied to select two SVM parameters for classification. Our work's aim is to optimize the SVMs by seeking for configurations of parameters which maximize the success rate and minimize the number of support vectors (i.e., two objetive functions). In the experiments, the performance of the search algorithms using a traditional random initialization was compared to the performance achieved by initializing the search process using the ML suggestions. We verified that the combination of the techniques with ML obtained solutions with higher quality on a set of 40 classification problems.

Published in:

Neural Networks (SBRN), 2012 Brazilian Symposium on

Date of Conference:

20-25 Oct. 2012