Evolving Complex Neural Networks Mauro Annunziato 1 , Ilaria Bertini 1 , Matteo De Felice 2 , Stefano Pizzuti 1 1 Energy, New technology and Environment Agency (ENEA) Via Anguillarese 301, 00123 Rome, Italy {mauro.annunziato, ilaria.bertini, stefano.pizzuti}@casaccia.enea.it 2 Dipartimento di Informatica ed Automazione, Università degli Studi di Roma “Roma Tre”, Via della Vasca Navale 79, 00146 Rome, Italy matteo.defelice@casaccia.enea.it Abstract. Complex networks like the scale-free model proposed by Barabasi- Albert are observed in many biological systems and the application of this topology to artificial neural network leads to interesting considerations. In this paper, we present a preliminary study on how to evolve neural networks with complex topologies. This approach is utilized in the problem of modeling a chemical process with the presence of unknown inputs (disturbance). The evolutionary algorithm we use considers an initial population of individuals with differents scale-free networks in the genotype and at the end of the algorithm we observe and analyze the topology of networks with the best performances. Experimentation on modeling a complex chemical process shows that performances of networks with complex topology are similar to the feed-forward ones but the analysis of the topology of the most performing networks leads to the conclusion that the distribution of input node information affects the network performance (modeling capability). Keywords: artificial life, complex networks, neural networks 1 Introduction Artificial Neural Networks (ANN) and Evolutionary Algorithms (EA) are both abstractions of natural processes. They are formulated into a computational model so that the learning power of neural networks and adaptive capabilities of evolutionary processes can be harnessed in an artificial life environment. “Adaptive learning”, as it is called, produces results that demonstrate how complex and purposeful behavior can be induced in a system by randomly varying the topology and the rules governing the system. Evolutionary algorithms can help determine optimized neural network architectures giving rise to a new branch of ANN known as Evolutionary Neural Networks [1] (ENN). It has been found [2] that, in most cases, the combinations of evolutionary algorithms and neural nets perform equally well (in terms of accuracy) and were as accurate as hand-designed neural networks trained with backpropagation [3]. However, some combinations of EAs and ANNs performed much better for some data than the hand-designed networks or other EA/ANN combinations. This suggests that in applications where accuracy is a premium, it might pay off to experiment with