A novel approach to automatic music composing: Using genetic algorithm

Artificial music composition is one of the ever rising problems of computer science. Genetic Algorithm has been one of the most useful means in our hands to solve optimization problems. By use of precise assumptions and adequate fitness function it is possible to change the music composing into an optimization problem. This paper proposes a new genetic algorithm for composing music. Considering entropy of the notes distribution as a factor of fitness function and developing mutation and crossover functions based on harmonic rules and trying to keep the melodies intact during these processes would result in a musical piece pleasant to human ears and interesting for human mind. This algorithm does not have the constraints of the previous algorithms. Restraining mutation and crossover functions with a goal of producing melodies based on acceptable melodies composed by humans, this algorithm is not bound to any genre, instrument or melody. The experimental results of this approach show that it is near to the human composing and the results produced from it are more acceptable than the ones produced by its predecessors