Music without Music Theory
July 6, 2017. . An article from EPFL news.The full article. (Indicated by ACM Technews).
Researchers at the Swiss Federal Institute of Technology in Lausanne (EPFL)
have developed a deep-learning algorithm that can generate melodies that imitate
a given style of music. The researchers say their "deep artificial composer"
(DAC) could one day generate convincing music for multiple instruments in real
The DAC, which generates new melodies that imitate the traditional folk music of Irish or Klezmer origin, produces musical scores of melodies, symbolic music written using notation, and does not generate audio files. "The deep artificial composer can produce complete melodies, with a beginning and an end, that are completely novel and that share features that we relate to style," says EPFL researcher Florian Colombo.
The system determines its own composition rules by extracting probability distributions from existing melodies using neural networks. The DAC is based on an artificial neural network, which already has proven useful for speech recognition.