Fig. 1.

Schematic of our evolutionary algorithm. Individuals are defined as connected subnetworks and are initialised using a depth first search algorithm; the selection path is represented with directed arrows starting from an initial node highlighted with a blue border (a). Optimisation is performed by applying mutation and crossover operators. The mutation operator randomly changes a node by maintaining network connectivity (b). The crossover operator merges two parent subnetworks, represented with green and yellow nodes, into a connected one and generates two new subnetworks applying a depth first search algorithm to an initial, randomly selected node; the selection paths generating the offspring are represented with blue and red directed arrows starting from initial nodes highlighted respectively with blue and red borders (c)