"Machine learning for live-coding improvisation"

Since 2015, Jason has been performing live improvised electronic music by livecoding cyclic traversals through a virtual space populated by audio samples organized by the t-SNE algorithm.  In this talk, he will discuss his hypotheses and observations over the last three years, and make the case for the validity of using dimensionality reduction to organize musical elements. Specifically, he will talk about curating and generating data sets, the wide range of possible features to describe audio and how they affect the output of the t-SNE algorithm, and different ways to be expressive with the structured data that results. The talk will include several demos and a short performance at the end.

Bio

Jason Levine is a Computational Artist and Musician who specializes in audio visualization, generative music,  livecoding, and interactive music systems. He has a Bachelor’s in Computer Science from Concordia University with a minor in Music Technology from McGill.  He is an alumnus of the School for Poetic Computation, a co-founder of Livecode.NYC, and in 2018 he was artist-in-residence at MIT. While Jason’s work crosses mediums, it is always dedicated to musical expression. He worked with Harmonix to build Kinect-controlled music interactions for Disney’s Fantasia: Music Evolved.  He worked with Ms. Lauryn Hill to create audio-reactive visuals for her 2017 tour. He created a suite of VR instruments that allowed users to remix Tokimonsta’s Grammy-nominated album Lune Rouge in the TheWaveVR. Most recently, he built an EMG-controlled audiovisual synthesizer for CTRL-Labs which was showcased at Slushfest 2018.  Jason is one half of the live electronic duo Scorpion Mouse, which has had the honor of performing at venues, conferences, and festivals (and a wedding!) in NYC, LA, Morelia, Guadalara, and Hong Kong. You can follow Jason’s code experiments here: https://www.instagram.com/livecodez/