Annotation

Annotation: This paper proposes an approach to improve the quality of Extreme Learning Machine (ELM) predictions using a genetic algorithm that implements an evolutionary process for neuron activation functions in the hidden layer of the model. To select the best candidate in the selection breeding process, both its computational complexity and the quality of the results obtained using it are considered. To validate the proposed approach, 4 different datasets are considered and used to evaluate the quality of different models for classification and regression tasks. Experimental results confirm that ELM model shows better results with activation functions obtained using Gene Expression Programming (GEP) algorithm than with classical activation functions used to solve similar problems.