A novel artificial bee colony optimization strategy-based extreme learning machine algorithm

Abstract

Extreme learning machine (ELM) is a kind of single-hidden layer feedforward neural networks (SLFNs). Compared with traditional neural networks algorithms, ELM is simpler in structure with higher learning speed and better generalization performance. Due to generating randomly input weights and biases of ELM, there can exist some non-optimal or unnecessary input weights and biases. In addition, ELM can need more hidden nodes which can make ELM respond slowly to unknown testing data. Consequently, a new NABC-ELM algorithm, which is optimized by a novel artificial bee colony called NABC, is proposed. To improve generalization performance of ELM, the NABC is applied to optimize input weights and biases. In NABC, the Tent chaotic opposition-based learning method is applied to initialize the population. Meanwhile, the self-adaptive search strategy is presented in the employed bee and onlooker bee phase. In addition, the Tent chaotic local search for scout bee is implemented. Finally, experiments on some popular classification data sets demonstrate that the proposed NABC-ELM can consistently get better generalization performance than some existing ELM variants.