I think of classical music as a more "canonical" target than atonal or dissonant music, though it depends on what community of researchers you're looking at. On

I think of classical music as a more "canonical" target than atonal or dissonant music, though it depends on what community of researchers you're looking at. One of the more famous researchers is David Cope, who spent much of his time on algorithmically modeling composition in the style of J.S. Bach. There's also an intermittent stream of papers on jazz improvization, e.g.: http://scholar.google.com/scholar?q=jazz+improvisation+icmc http://scholar.google.com/scholar?q=jazz+improvisation+icmc There is a good deal of avant-garde experimentation as well, in part because computer music has in many places/eras been larger in the field of music, rather than the field of computer science. Contemporary composers like Iannis Xenakis, Karlheinz Stockhausen, Pierre Boulez, etc., picked up and developed new technologies as they became available. But they were avant-garde composers interested in using these technologies as part of their compositional toolkit, not CS researchers interested in automating generation of music per se. One driving question for a certain group of composers is along the lines of: how can this new technology/technique help us make new, different music? Which leads to different experiments than the question: how can this new technology/technique help us automate the production of existing music? Sometimes that still ends up popularly oriented, e.g. in the '60s and '70s there was quite a bit of crossover between avant-garde synth experimentation and popular music, with Brian Eno as one hub in particular, and Moog synths finding their way onto all sorts of top-charting albums. Other times it's ended up less popularly oriented, like most of the granular-synthesis music.