Results on the Model Selection Game: Towards a Particle Swarm Model Selection Algorithm

H. Jair Escalante

National Institute of Astrophysics Optics and Electronics, Puebla, Mexico. hugojair

Many statistical pattern analysis methods and machine learning algorithms
have been proposed so far. These methods have been widely
used in many domains, including scientific, medical, and commercial
applications with great success. However, selecting the best method for
analyzing a determined dataset, is still a challenge; even a harder problem
is the estimation of the parameters for the selected algorithm. This
difficult problem is known as model selection. In this work we introduce
a novel algorithm for model selection based on particle swarm optimization
(PSO): particle swarm model selection, and preliminary results
from it are shown. The objective of our algorithm is to choose the optimal
model, from a pool of methods, for better generalization on unseen
data by casting the model selection problem to a PSO environment. Furthermore,
we describe results of our participation on the model selection
game 2006.