Algorithmic music composition [closed]

As it currently stands, this question is not a good fit for our Q&A format. We expect answers to be supported by facts, references, or expertise, but this question will likely solicit debate, arguments, polling, or extend开发者_C百科ed discussion. If you feel that this question can be improved and possibly reopened, visit the help center for guidance. Closed 11 years ago.

(edited) For anyone interested in music and artificial intelligence:

Do you know of any music-composing algorithm that produces really interesting, fun or intelligent music? And not something sounding like a random noise.

---

(Previous, too broad question:)

What are some state of the art (very good, non-boring) music composition algorithms, software, researches that you have heard of? Feel free to post any interesting link about this subject.

P.S. I don't mean programs that assist you at playing, but primarily anything that can compose melody by itself (or with little assistance).

OR: Analyses existing music pieces and tells how much it likes them :)

---

One of the leading researchers in algorithmic composition is David Cope of the University of California, Santa Cruz. His approach emphasizes machine learning, the results of which were impressively demonstrated in a 2006 performance.

http://www.wired.com/wired/archive/14.09/posts.html?pg=3

A good place to start would be with his aptly named book, The Algorithmic Composer, which covers much of his approach and provides most of the software he has written for his work.

http://books.google.com/books?id=rFGH07I2KTcC

Though not specifically algorithmic composition another invaluable resource is David Temperley's book, The Cognition of Basic Musical Structures, which provides quite a few models begging to be implemented.

http://books.google.com/books?id=IDoLEvTQuewC

Those two alone a pretty time consuming for anyone with an interest in that they are concrete enough that experimenting along the way is inevitable.

Hope that helps.

---

One possibility would be to use a hidden Markov model: feed it samples of music, and have it generate "similar" music.

One example: http://www.cogs.susx.ac.uk/users/christ/talks/music-making-with-HiMMs.pdf

I did something similar with Shakespeare's sonnets. The results were ... interesting. Amusing, at times.

---

• There's a search engine that lets you whistle a tune and that searches for music alike. I'm not sure whether http://www.midomi.com/ is what I originally heard of. You can for example play the music and see if it finds what you intended.

• A fellow student of me created a score composer for his Master's project. The input was humming or whistling and through FFT, music theory and combinatorial algorithms (I'm not sure whether it was simulated annealing). I'm not sure how it was related, but the project had something to do with the http://www.wikifonia.org/ project.

(edit)

• I heard a talk from someone who worked at http://last.fm. They analyze music (machine learning) as one of the ways to overcome the cold start problem in their recommender system. They try to predict how much a new song resembles other songs.