Towards designing neural network ensembles by evolution

This paper proposes a co-evolutionary learning system, i.e., CELS, to design neural network (NN) ensembles. CELS addresses the issue of automatic determination of the number of individual NNs in an ensemble and the exploitation of the interaction between individual NN design and combination. The idea of CELS is to encourage different individual NNs in the ensemble to learn different parts or aspects of the training data so that the ensemble can learn the whole training data better. The cooperation and specialisation among different individual NNs are considered during the individual NN design. This provides an opportunity for different NNs to interact with each other and to specialise. Experiments on two real-world problems demonstrate that CELS can produce NN ensembles with good generalisation ability.