You might find the work of Mason Bretan on this [0][1], and the video demo here <a href="https://www.youtube.com/watch?v=BbyvbO2F7ug" rel="nofollow">https://www.youtube.com/watch?v=BbyvbO2F7ug</a> relevant to your future work. Nice writeup!<p>[0] <a href="https://arxiv.org/abs/1612.03789" rel="nofollow">https://arxiv.org/abs/1612.03789</a><p>[1] <a href="https://arxiv.org/abs/1706.04486" rel="nofollow">https://arxiv.org/abs/1706.04486</a>
Nice write up!<p>I think you might see some interesting results using a non-parametric [one that doesn't require specifying the number of clusters apriori] clustering algorithm, like mean-shift. I've never seen an adaptation for discrete data like this, but it should be possible.<p>You could have the same tradeoff between note-distance and time-distance.
relevant: A Generative Theory of Tonal Music. <a href="https://mitpress.mit.edu/books/generative-theory-tonal-music" rel="nofollow">https://mitpress.mit.edu/books/generative-theory-tonal-music</a>