An Enhanced Online Sequential Extreme Learning Machine algorithm

In this paper, an Enhanced Online Sequential Extreme Learning Machine (EOS-ELM) algorithm for Single-hidden Layer Feedforward Neural Networks (SLFNs) with radial basis function (RBF) hidden nodes is proposed. The proposed EOS-ELM algorithm is an enhanced version of the OS-ELM of [8], which has been shown to be extremely fast with generalization performance better than other sequential training methods. The EOS-ELM algorithm adapts the node location, adjustment and pruning method of the MRAN of [3], so that the number of hidden nodes used in the OS-ELM can be modified. Simulation results show that the generalization performance of EOS-ELM is comparable to the OS-ELM and the number of nodes used by the EOS-ELM is reduced significantly.