Timothy K. Shih

An Internet-based Music Generation System using Reinforcement Learning

This study proposes a symbolic music generation system that integrates supervised learning and reinforcement learning and employs recurrent neural networks for sequence modeling, while a reinforcement learning module formulates music rules as reward functions to guide pitch, duration, and rhythm.