A Novel Artificial Bee Colony Learning System for Data Classification

Abstract

Training artificial neural networks (ANNs) is a common hard optimization problem. The process of neural nets training is generally defined on synaptic weights and thresholds of artificial neurons with the aim to find optimal or near-optimal values. Artificial bee colony (ABC) optimization has been successfully applied to several optimization problems, including the optimization of weights and biases of ANNs. This paper addresses the problem of feed-forward ANNs training by using a novel ABC variant named cooperative learning artificial bee colony algorithm (CLABC), which we have developed in our previous work. The performance of the CLABC-trained feed-forward ANN is validated on different classification problems, namely the XOR problem, the 3-bit parity, 4-bit encoder-decoder and Iris benchmark problems. The results are compared to other advanced optimization methods.