Enhanced Extreme Learning Machine with stacked generalization

This paper first reviews extreme learning machine (ELM) in light of coverpsilas theorem and interpolation for a comparative study with radial-basis function (RBF) networks. To improve generalization performance, a novel method of combining a set of single ELM networks using stacked generalization is proposed. Comparisons and experiment results show that the proposed stacking ELM outperforms a single ELM network for both regression and classification problems.