Determination Optimum SVMs Classifiers for Hyperspectral Imagery Based on Ant Colony Optimization
Referring to robustness of SVMs in high dimensional space, they are reliable tools for classification of hyperspectral imagery. However their performance is directly affected by two aspects: parameter determination and optimum feature subset selection. According to capacity of population based meta-heuristic optimization algorithm such as Ant Colony Optimization (ACO), they can find optimum or near optimum solution in complex optimization problems. This paper evaluates the potential of Binary ACO (BACO) in parameter determination, feature selection and both of them simultaneously in SVMs based classification system for hyperspectral imagery. Obtained results in comparison with genetic algorithm show superiority of BACO.
F. Samadzadegan and H. Hasani, "Determination Optimum SVMs Classifiers for Hyperspectral Imagery Based on Ant Colony Optimization", Key Engineering Materials, Vol. 500, pp. 792-798, 2012