A genetic algorithm for composing music (Q2853210)

The author presents an interesting application of genetic algorithm. The author uses genetic algorithm approach (GA) for generating music compositions. The encoding scheme and constructive elements of the GA are adapted to the problem of composing music. The goal is to meet some objective criteria of ``beautiful'' compositions to containing intervals that are pleasant to the human ear and to have a meaningful rhythm.NEWLINENEWLINE The composition is encoded by an array of tones and breaks with additional information about the length. In this way, an effective and quick control of the composition, tones and its rhythm is provided. This coding system enables the application of appropriate mathematical functions to tones, intervals and other ``musical'' parameters. The author implements modified genetic operators in order to enable significantly changing scheduling of pitches and breaks, which can restore good genetic material and prevent from premature convergence in bad suboptimal solutions. In the proposed GA, three types of mutation and selection are used, while crossover is omitted. The author thoroughly explains the reasons for omitting the crossover.NEWLINENEWLINE Finally, the author presents the analysis of GA solutions and experiments with the key parameters of the algorithm. The analysis of the obtained solutions in relation to the key parameters is also presented. However, the comparison of the GA results with the results of other algorithms for music compositions is omitted. It would be nice to evaluate the quality of the GA compositions in respect to the different techniques for algorithmic music composition, such as: knowledge-based systems, grammars, machine learning, fractal interpolation, stochastic binary subdivision, and many others.