Assisting students to compose music with deep neural networks and aesthetic measurement feedback

Razvan Paroiu, Stefan Trausan-Matu

Abstract: This paper introduces a novel approach that enables students to compose music using deep neural networks while simultaneously gaining insights into the functionality of various mathematically computed aesthetic measures. While there are existing applications that allow users to compose music with artificial intelligence, they often lack the feedback, which may be provided by newly developed mathematical aesthetic measurement techniques. Our approach bridges this gap by providing users with objective feedback for their generated melodies, fostering a better understanding of the underlying measurement methods. The application was evaluated by multiple participants, which made us aware of both the strengths and limitations of the application.