Léopold Crestel's PhD Defense

Edit

"Neural networks for automatic musical projective orchestration"

Edit

Content:
Orchestration is the art of composing a musical discourse over a combinatorial set of instrumental possibilities. For centuries, musical orchestration has only been addressed in an empirical way, as a scientific theory of orchestration appears elusive. Indeed, whereas harmony and counterpoint can rely on solid theoretical grounds, orchestration remains taught from collections of examples drawn from the repertoire. In this work, we address these questions within the machine learning framework, by proposing the first projective orchestration system. Hence, we start by formalizing this novel task. We focus our effort on projecting a piano piece (seen as an harmonic draft) onto a full symphonic orchestra, in the style of notable classic composers such as Haydn, Mozart or Beethoven. Hence, the first objective is to design a system of live orchestration, which takes as input the sequence of chords played by a pianist and generate in real-time its orchestration. Afterwards, we relax the real-time constraints in order to use slower but more powerful models and to generate scores in a non-causal way, which is closer to the writing process of a human composer. By observing a large dataset of orchestral music written by composers and their reduction for piano, we hope to be able to capture through statistical learning methods the mechanisms involved in the orchestration of a piano piece. Deep neural networks seem to be a promising lead for their ability to model complex behaviour from a large dataset and in an unsupervised way. More specifically, in the challenging context of symbolic music which is characterized by a high-dimensional target space and few examples, we investigate autoregressive models. At the price of a slower generation process, auto-regressive models allow to account for more complex dependencies between the different elements of the score, which we believe to be of the foremost importance in the case of orchestration. Léopold Crestel will defend his doctoral  carried out  in the <a href="/recherche/equipes-recherche/repmus/">Musical Representations</a> <a href="https://www.ircam.fr/recherche/equipes-recherche/systemes-et-signaux-sonores-audioacoustique-instruments-s3am/">team</a> (STMS - IRCAM/CNRS/Sorbonne Université/Ministère de la Culture). The jury will be composed: Philippe Esling, Sorbonne Université Jean-Gabriel Ganascia, Sorbonne Université Maarten Grachten, Independent consultant Florence Levé, Université de Picardie Jules Verne Yan Maresz, CNSMDP Geoffroy Peeters, Télécom Paristech Kamel Smaïli, LORIA, Nancy

Orchestration is the art of composing a musical discourse over a combinatorial set of instrumental possibilities. For centuries, musical orchestration has only been addressed in an empirical way, as a scientific theory of orchestration appears elusive. Indeed, whereas harmony and counterpoint can rely on solid theoretical grounds, orchestration remains taught from collections of examples drawn from the repertoire.

In this work, we address these questions within the machine learning framework, by proposing the first projective orchestration system. Hence, we start by formalizing this novel task. We focus our effort on projecting a piano piece (seen as an harmonic draft) onto a full symphonic orchestra, in the style of notable classic composers such as Haydn, Mozart or Beethoven. Hence, the first objective is to design a system of live orchestration, which takes as input the sequence of chords played by a pianist and generate in real-time its orchestration. Afterwards, we relax the real-time constraints in order to use slower but more powerful models and to generate scores in a non-causal way, which is closer to the writing process of a human composer.

By observing a large dataset of orchestral music written by composers and their reduction for piano, we hope to be able to capture through statistical learning methods the mechanisms involved in the orchestration of a piano piece. Deep neural networks seem to be a promising lead for their ability to model complex behaviour from a large dataset and in an unsupervised way. More specifically, in the challenging context of symbolic music which is characterized by a high-dimensional target space and few examples, we investigate autoregressive models. At the price of a slower generation process, auto-regressive models allow to account for more complex dependencies between the different elements of the score, which we believe to be of the foremost importance in the case of orchestration.

Léopold Crestel will defend his doctoral carried out in the Musical Representations team (STMS - IRCAM/CNRS/Sorbonne Université/Ministère de la Culture).

The jury will be composed:

Philippe Esling, Sorbonne Université
Jean-Gabriel Ganascia, Sorbonne Université
Maarten Grachten, Independent consultant
Florence Levé, Université de Picardie Jules Verne
Yan Maresz, CNSMDP
Geoffroy Peeters, Télécom Paristech
Kamel Smaïli, LORIA, Nancy

Edit