Using genetic algorithms for music composition: implications of early termination on aesthetic quality

Recently, the application of computer algorithms for music composition is increasing rapidly. Although music obtained from these algorithms may or may not be appealing, these algorithms produce optimal music based on the criteria defined by composers. Yet, studies that assess the relationship between the optimal music produced and the aesthetic quality of music is lacking. This study, therefore, seeks to compose monophonic music using genetic algorithm (GA) and assess how well it appeals to humans. Nine (9) melodies were composed using GA. The results show that GA performs best when terminated at ten thousand generations. The aesthetic quality evaluation conducted using an opinion survey demonstrated that compositions obtained using GA were best when terminated at 60000th generation. This suggests that although GAs may perform better in terms of composition time, it does not guarantee aesthetic music since there was no evidence that early convergence of GA models and will produce good aesthetics. It is therefore recommended that automatic composition methods must not focus solely on optimizing composition rules but reconcile empathy, intimacy, passion, and creativity to improve aesthetics.