Deep Learning for Music Generation - Course - UNIRIO - 2018 - Jean-Pierre Briot

Deep Learning for Music Generation - Course - UNIRIO - 2018 - Jean-Pierre Briot


More recent: Current course (1st Semester 2019)

Slides and programs

1. Introduction to computer music

Slides

2. Introduction to deep learning

Slides
MNIST handwritten classification (Code)

3. Music generation by feedforward

Slides
Chord accompaniment - value encoded (Code) Chord accompaniment - one-hot encoded (Code)
Chorale 3 voices counterpoint accompaniment (Code) Example of original chorale (Score) Example of original chorale (Audio) Example of generated chorale (Score) Example of generated chorale (Audio)

4. Music generation by decoding from an autoencoder

Slides
Melody generation - autoencoder (Code)
Example 1 of generated melody (Score) Example 1 of generated melody (Audio) Example 2 of generated melody (Score) Example 2 of generated melody (Audio) Example 3 of generated melody (Score) Example 3 of generated melody (Audio)
Pseudo digit generation - autoencoder (Code) Pseudo digit generation - stacked autoencoders (Code)

5. Music generation by iterative generation from a recurrent network

Slides
Melody generation - recurrent network (Code) Example of generated melody (Score) Example of generated melody (Audio)

6. GAN and compound network architectures

Slides

7. Reinforcement learning

Slides

8. Strategies for control of music generation

Slides

Proposed papers for presentation/evaluation

Papers

Presentations


Jean-Pierre.Briot@lip6.fr 31/10/2018