Analysis Of the Principle and Application of Machine Learning for Music Composition

Abstract

Contemporarily, machine learning and artificial intelligence are undergoing rapid developments in the field of arts. This paper explores the principles and applications of machine learning in music composition, tracing back to its inception in 1950s, and taking a brief look at the first works in the field of computer-generated music. It delves into the key principles of music composition using machine learning, and discusses the theory behind major models of recurrent neural networks (RNNs) and Generative Adversarial Networks (GANs) and how they are utilized in music composition. This paper also discusses about the real-world applications and potentials of current models, and finally analyze key limitations present in this field and points out a few directions for future works. This paper gives a brief and simple overview of the current status of machine learning in music composition, and provides an assistance to individuals interested in exploring this field and a valuable insight for future innovations.