This seminar proposes to display and discuss three research approaches to machine learning for music, which were adopted by the two speakers during the three years of their theses.
A first approach consists in studying machine learning for music with a scientific focus. Axel will present his thesis work, investigating Bayesian learning and information theory for audio analysis, raw generation, and synthesis space extraction from existing sound datasets.
A second approach consists in leveraging machine learning to design human musical interactions. Hugo will present his thesis work, relying on four human-centred methodologies to design (with) four learning algorithms for four data-driven interactive music systems.
A third approach consists in practicing with machine learning for music creation. Axel and Hugo will present their collaboration on aego, an improvisational piece with interactive sound and image, conceived with, and written for, one performer and one learning machine.