Learning Ensemble Performance for String Quartet by Practice

We introduce the ensemble practice of the string quartet. The music performance is studied recently, and most of the research aims at a computer and a human being playing together so far. On the other hand, the purpose of our research is to simulate the process that progress is made in the performance in accordance with the practice. Generally a player acquires performance through the practice. But, the research of simulating performance practice does not exist. We simulate the process that progress is made in the performance by the ensemble practice. Reinforcement learning is used for by this research. In the simulation, individual practice and ensemble practice are done by using `Kanon´ composed by Pachelbel. The player´s performance is getting better by the practice, and each player modifies the deviation of the tempo by ensemble practice to do after the individual practice