Book Image

Hands-On Music Generation with Magenta

By : Alexandre DuBreuil
Book Image

Hands-On Music Generation with Magenta

By: Alexandre DuBreuil

Overview of this book

The importance of machine learning (ML) in art is growing at a rapid pace due to recent advancements in the field, and Magenta is at the forefront of this innovation. With this book, you’ll follow a hands-on approach to using ML models for music generation, learning how to integrate them into an existing music production workflow. Complete with practical examples and explanations of the theoretical background required to understand the underlying technologies, this book is the perfect starting point to begin exploring music generation. The book will help you learn how to use the models in Magenta for generating percussion sequences, monophonic and polyphonic melodies in MIDI, and instrument sounds in raw audio. Through practical examples and in-depth explanations, you’ll understand ML models such as RNNs, VAEs, and GANs. Using this knowledge, you’ll create and train your own models for advanced music generation use cases, along with preparing new datasets. Finally, you’ll get to grips with integrating Magenta with other technologies, such as digital audio workstations (DAWs), and using Magenta.js to distribute music generation apps in the browser. By the end of this book, you'll be well-versed with Magenta and have developed the skills you need to use ML models for music generation in your own style.
Table of Contents (16 chapters)
1
Section 1: Introduction to Artwork Generation
3
Section 2: Music Generation with Machine Learning
8
Section 3: Training, Learning, and Generating a Specific Style
11
Section 4: Making Your Models Interact with Other Applications

Chapter 4: Latent Space Interpolation with MusicVAE

  1. The main use is dimensionality reduction, to force the network to learn important features, making it possible to reconstruct the original input. The downside of AE is that the latent space represented by the hidden layer is not continuous, making it hard to sample since the decoder won't be able to make sense of some of the points.

  2. The reconstruction loss penalizes the network when it creates outputs that are different from the input.
  3. In VAE, the latent space is continuous and smooth, making it possible to sample any point of the space and interpolate between two points. It is achieved by having the latent variables follow a probability distribution of P(z), often a Gaussian distribution.
  4. The KL divergence measures how much two probability distributions diverge from each other. When combined with the reconstruction loss...