Deep Learning with Audio

Overview

During the pilot period, we will develop free and open source teaching materials and code for the course, to make contemporary work in deep learning and audio more accessible to students who may not have advanced knowledge of programming or experience with setting up computational environments.

Students will learn the basic architecture of deep learning and artificial intelligence models for audio, work on course exercises and start building AI applications for their own purposes. The exercises and projects will be based on the Pure Data platform, which students are familiar with from previous courses. We will provide code templates that integrate the functionality from open source deep learning audio projects, such as Google’s Magenta, into Pure Data’s visual programming paradigm. We will also provide detailed setup instructions and automated scripts to make installation of the required tools as easy as possible (such as for Pure Data, Python, Conda, Magenta, PyExt).

The realisation of the pilot will improve students’ learning by involving them deeply in the implementation of their artistic and design ideas. Deep Learning with Audio is a project based course, where students will be working on specific projects, receiving instant and personalised feedback on their work. This inquiry-based hands-on experience, incorporating technology and literacy, will help students to learn the foundations of deep learning and procedural audio content generation, it will help them to understand how to build an AI, make a case study to determine how AI functions in music and sound design; utilising computational tools that perceive their own states and the state of the surrounding environment and are able to make decisions related to content generation processes. By the end of the course, students will know powerful ways to use advanced deep learning methods to manipulate and generate audio for sonic interaction as well as creating AI models for interactive music practices.