Advertisement

Deep Learning Techniques for Music Generation

  • Jean-Pierre Briot
  • Gaëtan Hadjeres
  • François-David Pachet
Book

Part of the Computational Synthesis and Creative Systems book series (CSACS)

Table of contents

  1. Front Matter
    Pages i-xxviii
  2. Jean-Pierre Briot, Gaëtan Hadjeres, François-David Pachet
    Pages 1-10
  3. Jean-Pierre Briot, Gaëtan Hadjeres, François-David Pachet
    Pages 11-13
  4. Jean-Pierre Briot, Gaëtan Hadjeres, François-David Pachet
    Pages 15-17
  5. Jean-Pierre Briot, Gaëtan Hadjeres, François-David Pachet
    Pages 19-49
  6. Jean-Pierre Briot, Gaëtan Hadjeres, François-David Pachet
    Pages 51-114
  7. Jean-Pierre Briot, Gaëtan Hadjeres, François-David Pachet
    Pages 115-222
  8. Jean-Pierre Briot, Gaëtan Hadjeres, François-David Pachet
    Pages 223-241
  9. Jean-Pierre Briot, Gaëtan Hadjeres, François-David Pachet
    Pages 243-249
  10. Back Matter
    Pages 251-284

About this book

Introduction

This book is a survey and analysis of how deep learning can be used to generate musical content. The authors offer a comprehensive presentation of the foundations of deep learning techniques for music generation. They also develop a conceptual framework used to classify and analyze various types of architecture, encoding models, generation strategies, and ways to control the generation. The five dimensions of this framework are: objective (the kind of musical content to be generated, e.g., melody, accompaniment); representation (the musical elements to be considered and how to encode them, e.g., chord, silence, piano roll, one-hot encoding); architecture (the structure organizing neurons, their connexions, and the flow of their activations, e.g., feedforward, recurrent, variational autoencoder); challenge (the desired properties and issues, e.g., variability, incrementality, adaptability); and strategy (the way to model and control the process of generation, e.g., single-step feedforward, iterative feedforward, decoder feedforward, sampling). To illustrate the possible design decisions and to allow comparison and correlation analysis they analyze and classify more than 40 systems, and they discuss important open challenges such as interactivity, originality, and structure.

The authors have extensive knowledge and experience in all related research, technical, performance, and business aspects. The book is suitable for students, practitioners, and researchers in the artificial intelligence, machine learning, and music creation domains. The reader does not require any prior knowledge about artificial neural networks, deep learning, or computer music. The text is fully supported with a comprehensive table of acronyms, bibliography, glossary, and index, and supplementary material is available from the authors' website.

Keywords

Music Generation Machine Learning Deep Learning Neural Networks Representation Artificial Intelligence Interactivity Music Creation Encoding Models Flow Machines

Authors and affiliations

  • Jean-Pierre Briot
    • 1
  • Gaëtan Hadjeres
    • 2
  • François-David Pachet
    • 3
  1. 1.LIP6, Sorbonne Université, CNRSParisFrance
  2. 2.Sony Computer Science LaboratoriesParisFrance
  3. 3.Spotify Creator Technology Research LabParisFrance

Bibliographic information

Industry Sectors
Pharma
Materials & Steel
Automotive
Chemical Manufacturing
Biotechnology
Finance, Business & Banking
Electronics
IT & Software
Telecommunications
Law
Consumer Packaged Goods
Energy, Utilities & Environment
Aerospace
Oil, Gas & Geosciences
Engineering