Incremental extreme learning machine

Abstract

This new theory shows that in order to let SLFNs work as universal approximators, one may simply randomly choose input-to-hidden nodes, and then we only need to adjust the output weights linking the hidden layer and the output layer. In such SLFNs implementations, the activation functions for additive nodes can be any bounded nonconstant piecewise continuous functions or the activation functions for RBF nodes can be any integrable piecewise continuous functions.We propose two incremental algorithms:1) Incremental extreme learning machine (I-ELM) 2) Convex I-ELM (CI-ELM).