Evolving ambience with Differential Evolution

Goal of the project

This repo contains the final project for the Global and Multi-Objective Optimization course @ DSSC. The main goal is to use Differential Evolution and variations on the theme to optimize some audio plugins in order to match a target guitar sound. The source code implements multiple DE algorithms with a Lamarckian take based on the papers cited in presentation.pdf

Folder content

The contents of the repo are the following:

• project_source.py: source code implementing the DE/rand/1 and JADE algorithm, together with a Lamarckian optimiziation step based on the Hookes-Jeeves optimization algorithm.
• example.ipynb: notebook containing an example of use of the algorithms
• songs: brief audio clips, divided into dry and wet, to test the algorithms
• presentation.pdf: overall peresentation of the project with focus on the algorithms used and the results obtained.

Requirements

To properly work, the source code needs Supermassive, Freq Echo and Space Modulator VSTs, distributed by Valhalla DSP. As for libraries, the needed ones are the following:

• numpy
• pedalboard
• scipy
• librosa