“…In the present work, a genetic algorithm is used to iteratively search for compositions that optimize desired properties as predicted by the neural-network model, both with constraints, optimizing existing alloys, and without constraints, freely searching for novel alloys. GAs have been established for some time as a powerful tool for materials science, 19,20 in particular being frequently used to optimize the properties of existing materials, 21,22 the parameters of processing techniques, 23,24 and to design entirely novel materials. 25,26 GAs have also been applied in the search for novel glass-forming alloy compositions; Bansal et al 27 applied a NN and GA to identify glassy structures in Cu-Zr alloys with good resistance to shear deformation, Sun et al 28 used a GA to search for energetically favoured packing motifs in glassy Cu-Zr and Al-Sm alloys, and Tripathi et al applied genetic programming to identify glass-forming ability criteria.…”