Fine-Tuning of Support Vector Machine Parameters using Racing Algorithms
Pericles B. C. de Miranda, Paulo Ricardo da Silva Soares, Ricardo B. C. Prudêncio · The European Symposium on Artificial Neural Networks · 2014
This paper investigates the iterative racing approach, I/F- Race, for selecting parameters of SVMs. As a racing algorithm, I/F-Race eliminates candidate models as soon as there is sucient statistical ev- idence of their inferiority relative to other models with respect to the objective. The results revealed that the I/F-Race algorithm was able to achieve better parameter values in comparison to default parameters used in literature and parameters suggested by particle swarm optimization techniques.