ELM Feature Mappings Learning: Single-Hidden-Layer Feedforward Network without Output Weight

Abstract

According to conventional neural network theories, the feature of single-hidden-layer feedforward neural networks(SLFNs) resorts to parameters of the weighted connections and hidden nodes. SLFNs are universal approximators when at least the parameters of the networks including hidden-node parameter and output weight exist. Unlike above neural network theories, this paper indicates that in order to let SLFNs work as universal approximators, one may simply calculate only the hidden node parameter and the output weight is not required at all. In other words, this proposed neural network architecture can be considered as a standard SLFN without output weights. Furthermore, this paper presents experiments which show that the proposed learning method tends to extremely reduce network output error to a very small number with only several hidden nodes. Simulation results demonstrate that the proposed method can provide hundreds times faster learning speed compared to other learning algorithms including BP, SVM/SVR and other ELM methods.