Title:
Data-driven sound and interaction design

Abstract:
The digital luthier crafts sound models that a) access a particular space of sound, and b) provide an interface for navigating that space in a particular way. The traditional approach to modeling by programming signal processing algorithms is labor intensive, and rarely results in sounds that are as rich and complex as those in our natural physical world. Deep learning techniques enable a data-driven approach to sound modeling. The designer provides a collection of sounds as a way of defining the domain of the desired instrument, and the system organizes them in a "space" defined by latent factors. Navigating is a matter of choosing paths between points in the learned sound space which is intuitive and gives the designers a flexible way of building musical expressivity into their instruments.
In this talk, I will discuss some of the deep learning tools available to instrument makers and how they can work together as part of a "sound model factory" for creating interactive instruments from sound collections.

Bio:
Before joining MTG in August of 2022, Lonce Wyse was an Associate Professor in the Department of Communications and New Media at the National University of Singapore where he also founded and directed the Arts and Creativity Lab within the Interactive and Digital Media Institute. He holds a PhD in Cognitive and Neural Systems (Boston University, 1994), and was a Fulbright Scholar the following year at Taiwan National University. He serves on the editorial boards of the Computer Music Journal (MIT Press), Organized Sound (Cambridge University Press), and the International Journal of Performance Arts and Digital Media. Research topics include developing neural networks as interactive sound synthesis models, sound perception, real-time musical communication and notation, and networked music making. He has taught courses on sonic arts, sound and interaction, software studies, creative coding, and media art.