Abstract
This paper presents an extension of our earlier work [12] on single-layered complex-valued neural network (CVNN). In the earlier work, we proposed a new class of activation functions for complex-valued neuron (CVN) in a view to solving real-valued classification problems. To improve the performance, we investigate the ensemble of single-layered CVNNs in this paper. We applied two ensemble methods-bagging and negative correlation learning, to create the ensembles. Experimental results on a number of real-world benchmark problems show a substantial performance improvement over an individual single-layered CVNN classifier, and thus justify the application of CVNN ensembles on the classification problems.
