Would it be possible to make a statistical model of a jazz solo and use it to create new ones? Take a standard, and let’s focus on the saxaphone, say. Go to the solo and estimate a Markov transition kernel which tells you the probability distribution over the next tone conditional on the previous tone. In particular you want the joint probability distribution over the following note (or just the interval) and the note’s value (eighth, quarter, etc.) Feed it tons and tons of recordings of sax solos for the same tune (that’s why you want a standard.)
Once you have estimated your kernel, simulate it. Will it be music? How much of an improvement do you get if the state variable is the last two notes instead of just one? If your state variable is the last n notes, at what n are improvements no longer noticeable?
Cheap Talk 
15 comments
Comments feed for this article
May 7, 2013 at 1:34 am
Anonymous
https://news.ycombinator.com/item?id=4709472
May 7, 2013 at 2:08 am
Loren
http://www.hooktheory.com/blog/i-analyzed-the-chords-of-1300-popular-songs-for-patterns-this-is-what-i-found/
The linked post gives you an explanation of the site, then you can click on trends to see the typical chord progression in popular songs. This obviously isn’t the same as jazz, which is more free-flowing and I guess void of rules than pop, but it’s a similar idea.
May 7, 2013 at 7:18 am
Angry Dude
Wild idea. DARPA is starting a new program Investigating probabilistic programming languages. They will be creating a set of test problems. Perhaps this could be one?
June 10, 2013 at 2:16 am
Jan
found this at TOD in the comments:Bill R.June 11, 2012 at 11:16 amI’m from Oregon. Grew up in Southern Oregon in a tibmer economy now largely gone. I will tell you how private equity companies like Bain ruin an economy. Remember, another word for private equity is “corporate raider.” In the Medford, Ore. area of Oregon there was a very successful tibmer company called MedCo. They had several mills and employed thousands of workers in high paying jobs. They owned millions of acres of tibmer growing company. They prided themselves in a policy of sustained yield in their harvest of tibmer. That way a supply of tibmer would always be there for their mills. Along came a private equity firm of corporate raiders, investors from Texas. They did a leveraged buyout of MedCo. In other words they did a hostile take over of a well functioning company that did not want to be bought out. They then proceeded to close all the mills, fire all the workers, loot the pension funds, and cut all the tibmer, shipping it overseas. These wealthy Texas raiders destroyed a progressive company that had functioned for generations as a job base for an entire region of Southern Oregon. Entire communities were thrown into poverty and unto government safety net programs. Lands were sold for real estate development with attempts to bypass land use laws. (I know because my father was running for mayor of Medford and was offered financial support if he would support their move to fire a city manager who insisted on enforcing land use laws.) So not only did this private equity company destroy and economic base but they corrupted the political process and public officials. My father lost the election, and his opponent won with the support of this private equity firm. This private equity firm left the government with a toxic waste dump to clean up and thousands of ruined lives to restore and laid waste to an entire forest. This is the legacy of MItt Romney and the GOP. This is their plan for America.
June 13, 2013 at 9:45 am
binagnj
CGcAbt lminvxllqbpj
June 13, 2013 at 9:45 am
jfvgrx
xnIFd1 eofzgmsbirih
October 17, 2013 at 10:00 am
Regina
I couldn’t agree more about gecrrtaiis. I miss sitting with some of the elderly individuals I used to help and they would tell me stories for hours about the things that they witness when they were younger, and the lives that they lived up to the point that they needed care. I have always felt that the elderly know how to live their lives better than anyone, and they for sure never take it for granted! Sometimes the people taking care of them are not good people though, and it always breaks my heart to see someone talking badly to an elderly man or woman.I like how you made a switch from little tiny babies that have no real experience with life, to elderly men and women who have lived life to its fullest and still have more life to share with others. Such opposites!
October 17, 2013 at 12:39 pm
Buff
here
May 7, 2013 at 10:37 am
Daniel Bernardes
I saw a conference last year from a physicist and musician on the topic (http://eccs2012.eu/public_session.php) and in his research he found there are long-range correlations in classical music (and jazz if I’m not mistaken), which would imply that short-range correlation models, such as Markovian models, would be unfit to generate “human-like” synthetic jazz…
May 8, 2013 at 12:05 am
E
There must be some fancy techniques out there to estimate approximate periodicities? If so, I’d imagine there’s some great combination of that with the model you describe.
May 8, 2013 at 8:44 pm
Charlie
In my opinion as a jazz musician the state space needed is huge. I think it would include things you haven’t thought about, like the history of other players actions and information about upcoming song characteristics (the notion of “upcoming” is itself a state variable given you are choosing the length of the note). If you really want to do it right it would include your best guess of other players notes at time t, t+1, t+2, etc. For instance my best guess of the voicing the piano player will use on the next chord. Also, you need to add (at a minimum) dynamics and and note style to the state space. Although on many Coltrane solos in the late 50’s and early 60’s dynamics would not be a state variable.
May 9, 2013 at 9:08 pm
Brady Hauth
This method has been done.
May 21, 2013 at 1:02 pm
tPlusOne
[…] https://cheaptalk.org/2013/05/06/blogging-something-i-know-nothing-about-8/ […]
May 22, 2013 at 1:12 pm
oaw
Richard C. Pinkerton 1956 Published in Scientific American.
(Quick Google Search!)
June 10, 2013 at 3:39 pm
Mona
I have already slpeeing on my Honda Jazz 1.4 100ch version 2006. A lot of space.It’s a good car with good engine. But her practical side is not good, it’s really hard to sleep in it well, a kind of ferrari. Especially when it is large in size. I bought it to get from point A to B. It works, it’s good crap. Good Continuation. Jp