Evolving artificial neural network structure using grammar encoding and colonial competitive algorithm

Evolving artificial neural network usually refers to network structure evolution leaving the network's parameters to be trained using conventional algorithms. In this paper, we present a new method for artificial neural network evolution that evolves the network structure along with the network parameters. The proposed method uses grammatical encoding together with colonial competitive algorithm to evolve artificial neural network structure and parameters. This allows for a better description of the network using a formal grammar allowing the network architecture to shape the resulting search space in order to meet each problem requirement. The proposed method is compared with other five methods for artificial neural network training and is evaluated using four known regression problems. In all four datasets, the proposed method outperforms its competitors.