A Meta-Instrument for Interactive, On-the-Fly Machine Learning

Fiebrink, Rebecca and Trueman, Dan and Cook, Perry R.

Proceedings of the International Conference on New Interfaces for Musical Expression

Supervised learning methods have long been used to allow musical interface designers to generate new mappings by example. We propose a method for harnessing machine learning algorithms within a radically interactive paradigm, in which the designer may repeatedly generate examples, train a learner, evaluate outcomes, and modify parameters in real-time within a single software environment. We describe our meta-instrument, the Wekinator, which allows a user to engage in on-the-fly learning using arbitrary control modalities and sound synthesis environments. We provide details regarding the system implementation and discuss our experiences using the Wekinator for experimentation and performance.