Entropy as measure for creativity


E =  − ∑ipilog pi

For each midi I consider the information contained in rhythm, chromatism and the total one. I compute the histograms relative to each single note duration and picth, separately and jointly.

Song rhythmic cromatic total
Flutto 2.024 5.239 6.56

I compute the relative distances between songs:


dij2 = (Eir − Ejr)2 + (Eic − Ejc)2 + (Eit − Ejt)2

and build the graph in neo4j for those songs whose distance is inferior of 0.5.

CREATE (Entropy:Measure {label:'measure for creativity'})
CREATE (Aere:Song { id:'0', name:'Aere', entropy:'4.962'})
CREATE (AjdeJano) - [:DIST{d: 0.470}] -> (Prolecic),

distance songs distance between songs

You can see a clear connection to the songs which is pretty much the one we expect.

The evolution of entropy depends on the author and stabilizes over time, in same cases it shrinks.

entropy evolution evolution of entropy

In the graph Bach’s ciaccone, lazy bird of Coltrane and Vitali ciaccona where analyzed.