Implement Music Generation with GAN: A Systematic Review

Music generation has a long history, which can be a tool to decrease human intervention in the process. Recently, it is widely achieved to generate mellifluous music based on generative adversarial network (GAN), which is one of the deep learning models on unsupervised learning. One of the advantages of GAN is that it uses generative model and discriminative model to learn mutually with more realistic and higher accuracy. In this review, we focus on the overview achievement with GAN to generate music. Specifically, the definition and GAN methods are introduced first. Subsequently, the application in music generation as well as the corresponding drawbacks are discussed accordingly. These results will offer a guideline for future research in music generation with machine learning techniques.