Tutorial Structure

Running the notebooks

---

The practical parts of this tutorial are delivered via Google Colab notebooks, found in:

• Notebook 1 - Introduction to Wavelets

• Notebook 2 - Scattering Transforms with Kymatio

• Notebook 3 - Musical Playing Technique Classification

• Notebook 4 - GEAR: Generative Evaluation of Audio Representations

• Notebook 5 - DTFA: Differentiable Time-Frequency Analysis

Notebook 3 - Musical Playing Technique Classification makes use of a dataset of musical instrument playing techniques with precomputed features. To access this dataset, please see this tutorial’s dedicated channel on the ISMIR23 Slack workspace, or contact the Authors.

Installation

---

We rely on the accessible GPU acceleration that’s provided in the Colab environment. All installation is handled within the Colab notebooks. We make use of a supplementary package s1dt for utility functions, feature extraction, and visualizations: ismir23-kymatio-tools. The main dependencies in each Colab notebook are kymatio-0.4-dev and s1dt, which can be installed from source:

!pip install git+https://github.com/kymatio/kymatio.git # install kymatio v0.4-dev from source
!pip install git+https://github.com/cyrusvahidi/ismir23-kymatio-tools.git # install helper library for this tutorial

Community

---

Get in touch via the Kymatio community on Github or X. You can also contact the Authors.

Note

The contents of this book will be continually updated after the ISMIR tutorial on November 5th 2023. Time-frequency scattering is currently only available in beta via the kymatio.torch frontend. It can be installed from source and will be officially released in Kymatio v0.4. Stay up-to-date on X.