Musical synthesis for certain music styles based on machine learning algorithms

Abstract

. The machine learning sphere is one of the most important and growing science spheres nowadays. Many algorithms dealing with images exist, but attention to music is growing as well. Some algorithms classify music by genre while others developed recently deal with music synthesis. The goal of this paper is to present an algorithm, which syntheses a music audio file on the basis of two certain fragments of different music compositions. This paper provides a theoretical basis of neural networks and convolutional neural networks in particular, describing how to construct a working model to create an audio output from two inputs, and shows how the network parameters can be evaluated and changed to fit the model better and generate better results.