Publications and Research
Document Type
Article
Publication Date
9-3-2026
Abstract
Generative systems that react to live musicians require rapid analysis of musical data, which rules out deep learning models: they cannot be trained within the time constraints of live performance. But because analysis results are often transformed before use, we are free to reduce the parameters that undergo transformation to a small set of primitive states. We address this coincidence of constraint and opportunity with an algorithm for online discovery of maximal musical motives that achieves speed through lossy compression: the pitch and inter-onset-interval deltas for all pairs of events in a potential motive are reduced to two-bit values, conceptualized as the edges of a complete graph. Column-major bit-packing of edges into register-sized words makes per-note motive discovery cheap enough to maintain the maximal set incrementally, so that a query reads results rather than computing them.
Comments
This paper was originally presented at the Proceedings of the 7th Conference on AI Music Creativity (AIMC 2026), Berlin, Germany, available at https://doi.org/10.5281/zenodo.22278396
This work is distributed under a Creative Commons Attribution 4.0 International License (CC BY 4.0).