Abstract
In this paper, an improved particle swarm optimization algorithm is proposed to train the fuzzy support vector machine (FSVM) for pattern multi-classification. In the improved algorithm, the particles studies not only from itself and the best one but also from the mean value of some other particles. In addition, adaptive mutation was introduced to reduce the rate of premature convergence. The experimental results on MNIST character recognition show that the improved algorithm is feasible and effective for FSVM training. © 2008 Springer-Verlag Berlin Heidelberg.
Cite
CITATION STYLE
Li, Y., Bai, B., & Zhang, Y. (2008). Fuzzy SVM training based on the improved particle swarm optimization. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 5227 LNAI, pp. 566–574). https://doi.org/10.1007/978-3-540-85984-0_68
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
