The Study on Initialization Aspects of the Extreme Learning Machine Parameters by Random Values

This paper investigates different ways of generating random numbers to initialize the parameters of Extreme Learning Machine (ELM) model. For this purpose, uniform and normal distribution of a random variable is considered. The models are compared in the context of solving a regression problem, for which synthetic datasets of different complexity are generated using a multilayer radial basis function network (RBF network). The results obtained from the model comparison are analyzed and discussed. The conclusion suggests further steps for the development of the topic discussed in this paper.