Two Parts Inventions' Auto Composition Base On Genetic Algorithm And Neural Networks

Algorithmic Composition (sometime also referred to "automated composition") basically refers to "the process of using some formal process to make music with minimal human intervention" (Alpern, 1995). GA(Genetic Algorithm) Composition and NN(neural networks) Composition basically refers to use GA and NN technology to make music with minimal human intervention. Because of the complexity of music composing , how to set the fitness function and design the training samples become a very important problem when we use these technology to make music.The motive is always considered as a core segment in polyphonic music. And counterpoint is one of very important skills on polyphonic music making. In this paper firstly it introduced the concept of GA and NN, summarized the researching status quo and developing direction on machine composition and expounded the making patterns by using these two technologies. Then it creates a module of Bach's two part inventions by researching the special configuration of inventions. During the making process it sets the fitness function based on the rules of the music knowledge of the motive making firstly, and use the GA to make new motive segment from the random notes. Then create the training samples based on rules of counterpoint skills, use these samples to train the NN and make counterpoint segments. Finally, combine these new segments in order to get a new two part inventions.The destination of our research consists in enriching the methods of algorithmic composition and developing the digital music technology. And these have important academic significance and application value.