GA-based Music Arranging for Guitar

In this paper we describe a system for converting a piece of music into an appropriate arrangement for guitar. The system is supported by previous research into the generation of guitar performance information (tablature) with genetic algorithms. This technology is a significant part of an evaluation function that is designed to differentiate between candidate arrangements on the basis of both faithfulness to the original piece and on playability. We describe both a genetic algorithm and a greedy hill-climber that use this function to find good arrangements. Both techniques have demonstrated an ability to create viable arrangements of the pieces tested.