As some of you may or may not know, I spent last weekend away at a conference. The upshot was probably getting into the top eight in the competitive programming contest, but grading the poster conference was also fun. Anyway, one poster in particular was pretty awesome. I didn't have the foresight to photograph it, but I will retype the blurb from the booklet and list off the conclusions that I remember.
Quote
ENTROPIC MODELS OF MUSIC: USING INFORMATION THEORY TO ANALYSIS[sic] MUSIC
Author: Ian Jones, SUNY Plattsburgh
Advisor: Delbert Hart, SUNY Plattsburgh
A piece of music has many different characteristics that make that piece unique and recognizable. The fundamental building blocks of music are melody, harmony, and rhythm (Simon 2007). Being able to break down a song into its piece is often an arduous task. Previous work by Watts (1974, 1979), Kerns (2001), and Simon (2006, 2007) have developed adaptations of Entropy equations into forms suitable for music. There are four main equations, Melodic Entropy, Harmonic Entropy, Rhythmic Entropy, and Composite Musical Entropy.[sic]
Claude Shannon developed techniques to describe the limits of compressison and reliability of data whilst being transmitted. The main measure in Information Theory is entropy. Entropy H, is defined as a discrete random variable X and the measure of the amount of uncertainty associated with the value of X.[sic]
Music is clearly a form of information. It can be as complicated a message as a Beethoven Symphony or as simple as a lone vocalist. The question exists is what is the information content of these different composition and aside from their musical complexities, is there greater entropy in a Beethoven piece versus that lone vocalist.[sic]
This study will look at the entropy values of various pieces of music in the form of guitar tablature from modern pop to classical pieces, calculate entropy values for the fundamental pieces of the musics and be able to quantify the music by its entropy values.[sic] If assuming music as a language, and a base line entropy value is established, then an analysis of a particular song ccan be achieved.[sic] Based upon entropy and other factors such as sales and chart positions, it can be determined if popularity correlates to a divergence from the average entropy values. Concisely, can the entropy of a song point to how well a song will be received.[sic]
Aside from Mr Jones' problems with the English language, the poster looked promising. He graphed the entropy of every piece in the data set, and graphed a line at the average entropy position. His results were that classical music had the highest entropy and pop music the lowest. Of the pieces in the data set, Michael Jackson's 'Beat It' had the lowest entropy -- meaning it was the most predictable.
Who the hell would enjoy entropic, unpredictable music?
me, if it is not too extreme.
Quote from: Sigmatic on April 20, 2010, 09:12:20 PM
Who the hell would enjoy entropic, unpredictable music?
If I recall correctly, the piece with the greatest entropy was by Bach.
I might not be recalling correctly, though. I don't think Bach wrote for the guitar.
That would be a "no." :lol:
Though I think he had some compositions for the lute?
Bach's music was very mathematical
Well... Music isn't going to be any good on either extreme end of the spectrum. The minimum entropy would be a pure square-wave tone that doesn't change (I don't know what the minimum in guitar tab form is). The max would be white noise (again, I dunno what the guitar tab equivalent is). Everything worth listening to is in a very small segment quite near the middle.
I mostly found it amusing that a grad student was using information theory to take a dig at Michael Jackson.
It depends purely on taste also Mr Enki :) Some enjoy the disoragnised sounds, others can only handle 4 notes for 3 minutes.
It is why my favourite style is Classical, it is always different to listen to (Of course sometimes if you hear one from the same composer [Or sometimes the peformer who doesn't change their technique], you've heard them all) ~ Of course most Classical I listen to isn't Disorganised it does follow an intricate pattern
Quote from: Sigmatic on April 20, 2010, 09:12:20 PM
Who the hell would enjoy entropic, unpredictable music?
The word "entropy" is jargon in information theory. Context clues, son.
Low entropy would be singing "this is the song that never ends [...]" forever. The lowest possible would be silence or singing "la" every second, forever, always at the same pitch. Or silence. Idk, really.
Do you want to listen to "99 bottles of beer on the wall", all the way to zero?
It's kinda a question of aneristic vs eristic delusions--neither is particularly healthy or fun for very long.
Alright, my bad. But please refrain from calling me "son", unless you happen to be at least twice my age.
So what's the point of studying entropy levels in music? Is it going to make better music, or is it just an exercise in measuring entropy?
It is probably going to somehow relate to human psychology by my guess... I am hoping it is just an exercise in measuring the Entropy however.
How would the focus of Entropy on music help people produce better results?
"Make it sound a bit more Disorganised, just turn the knob there"
Quote from: Sigmatic on April 20, 2010, 09:12:20 PM
Who the hell would enjoy entropic, unpredictable music?
WHY DO YOU HATE SQUAREPUSHER
http://www.youtube.com/watch?v=RBgn8QxYB6U
http://www.youtube.com/watch?v=BxEb2FrQUbE [track starts at 3:12 in]
also "Beat It" is a pretty swote track too, IMO:
http://www.youtube.com/watch?v=Ow-JyZ9xezk
btw I would think Squarepusher or Venetian Snares (FUCK HIM!!! :argh!:) would probably beat Bach in entropy hands down. In fact, so would any freestyle modern improv jazz wanking, since it's basically tonal whitenoise anyway :-P
BTW this is actually a very interesting topic, on the subject of order vs disorder.
Cause on the one hand, you have the lowest entropy, which is hitting the same note over and over. This always sounds pretty much the same.
And on the other end, there is highest entropy, which would be like throwing a d12 for every note and picking a random halfnote from an octave, or something. If you ever listened to purely randomly selected notes, you will have noticed that this ALSO always sounds the same! Well, you know that of course technically it never is exactly the same, but to your brain it's "that jumbled mess of 12 notes" every time. The chances of getting any melodic structure are very very slim.
If you look at my thread in Bring & Brag "Data Driven Design Doodles", one of my posts there talks about a similar problem, except not for sound, but for drawing shapes.
The idea is the same. Draw shapes that are too simple and they always look pretty much the same. Draw shapes that are too complicated and they always look like a big jumbled mess of lines.
I've tried a stab at random music generation myself, unfortunately I'm not even very good in making non-random music, so it turned out a bit simplistic (and I had a lot of help from a friend about harmony and chord progressions and shit). The end result is pretty funny, the melodies are random, but they always carry the same sort of "feel" or "emotion" because it selects from a limited set of chord progressions.
What happened to me was that while programming it, I would hear the melodies over and over again, and they would stick in my head! Except, there was no fixed melody to be stuck in my head, so I found myself humming, trying to hum something that I just couldn't put my finger on. It was an odd sensation.
I have to say, I did manage to program a pretty kickass randomized drum/percussion rhythm. I suppose I just have more feeling for that then, I just wrote a bunch of rules that seemed "right" to me, and out came some surprisingly funky shit, especially with some tweaking. For the notes I just had the advice my friend was giving me.
for anyone interested, the random music generator is here BTW: http://ritz.no.sapo.pt/TUTE0.EXE yeah it's an EXE file so run it at your own risk. I wrote it, so it's not a virus. Unless those bitches at no.sapo.pt switched it for one, depending on how paranoid you are (it's a random free portuguese webhoster I signed up for using an online translator many years ago, I have no idea who they are).
Quote from: Sigmatic on April 20, 2010, 11:08:40 PM
Alright, my bad. But please refrain from calling me "son", unless you happen to be at least twice my age.
So what's the point of studying entropy levels in music? Is it going to make better music, or is it just an exercise in measuring entropy?
Will do. Don't know quite why I did that bit anyways.
Quote from: Triple Zero on April 20, 2010, 11:35:23 PM
I've tried a stab at random music generation myself, unfortunately I'm not even very good in making non-random music, so it turned out a bit simplistic (and I had a lot of help from a friend about harmony and chord progressions and shit). The end result is pretty funny, the melodies are random, but they always carry the same sort of "feel" or "emotion" because it selects from a limited set of chord progressions.
What happened to me was that while programming it, I would hear the melodies over and over again, and they would stick in my head! Except, there was no fixed melody to be stuck in my head, so I found myself humming, trying to hum something that I just couldn't put my finger on. It was an odd sensation.
I have to say, I did manage to program a pretty kickass randomized drum/percussion rhythm. I suppose I just have more feeling for that then, I just wrote a bunch of rules that seemed "right" to me, and out came some surprisingly funky shit, especially with some tweaking. For the notes I just had the advice my friend was giving me.
Oooh, generated music!
Ever checked out http://tones.wolfram.com/ ?
I think your link is giving me a 404, but I don't speak portugese.
There is also, by the way, darwintunes. I also read about a program that writes music mimicing the styles of other music input -- kind of like a markov chain chatbot, except for music, and (probably) without using a markov model.
Quote from: Template on April 20, 2010, 11:39:17 PM
Ever checked out http://tones.wolfram.com/ ?
no, will check that out later.
Hm ENKI it really works for me, try it with referrers off?
Trip 0, I like this program you wrote. Even though the notes are random, the form is VERY mathmetical, it actually made some nice sounding results. The thing that is achieving unison here is the synth (Or my focus is just retarded, the obvious one would be drum also) .. I think without it, it would sound totally discord ~ though I think after a certain point you can pull it off without the synth.
Of course near the end sometimes it feels the piece was concluded, sometimes it doesn't - but I think to alleviate this, make it more biased into playing the first note that was heard from the instrument (And perhaps throw it in a chord or something) ~ and something that is longer than a 8th or a 16th
What language did you write this in?
Hahaha I can't stop listening to it, its kinda growing on me ... It sounds futuristic!
It was C++ with DirectX. It's been a loooooooong time, I dunno if I could do it again (in C++, I mean).
Also all of the sounds you hear are from a synth :) I wrote the software synthesizer that is generating the waveforms (it doesn't use sampled or recorded sounds) myself as well.
I presume you mean the ... I think they call it a "drone" or a "pad" in synthesizer terms. You get the sound by adding up a bunch of slightly detuned sawtooth waves, makes a nice typical indeed very synthethic filler sound. As well as probably a lowpass resonance filter and some simple reverb (which probably were just one or two echos layered on eachother--much easier than coding an actual reverb).
Um but that has nothing to do with random music generation :) I could have used MIDI sounds as well, it's just that I love playing around with synthesized sounds and programming weirs effects (as long as they dont require me to write any melodies, I am fine)
It worked when I grabbed it with wget. Thanks :-)
http://darwintunes.org/ and http://artsites.ucsc.edu/faculty/cope/experiments.htm are what I was thinking of, btw.
http://www.youtube.com/watch?v=Yt_tjQSdl30
Oh! The way I heard it was in 3 layers
1) Drums
2) Keyboard Synth
3) Electric/Bass Guitar/Another Keyboard
But that is very interesting how you approached the waveforms! That is a pretty smooth synth you wrote there
Are you going to be mucking around with anything in the near future?
There's the drums, which are three layers, one is the hard hitting bits of the bassdrum and "snaredrum" (quotes because synthesizing a proper snaredrum is hard), second is a bunch of random percussive and bass sounds (half of which aren't even based on actual drum instruments, but is just sound stuff I thought would fill out the rhythm). Third is that pretty much constant stream of hihats.
Then there's the keyboard synth, which are the long drawn out "synth" notes, which always play the three (?) "important" notes in the chord.
Then there's the "bassline", which is a random percussive melody of, well, those bwob bwob bwob bassline sounds. I think I coupled the rhythm to correspond a bit with the drumline. It consists of two or three notes from the current chord.
And finally there's the "lead tune", which is the higher pitched keyboard sound. It picks notes from the chord and a few in between that my friend told me would also "work".
This is all from memory btw, my gf is asleep so I can't play it. And the original code is on those old harddisks that I hope still work after the fire, but I don't really have a computer I can plug them in to.
And no, see I used to be a demoscene programmer. Wrote 4096 byte demos. The last one I made actually had a tiny softsynth in it. But I quit the scene in 2000 I just didn't feel like it anymore and went "well it's just a hobby" and quit. [what was actually the cause of this, I am still trying to figure out. Actually I'm desparately trying to get back to that mindstate]
Either way I'm not going to code C++ anymore :) But that's just because I fell in love with Python :-P Unfortunately Python is not fast enough to do real time sound synthesis.
My current programming projects do not have much to do with sound. There's a Forum-bot I'm trying to write, a thing that tracks and graphs WiFi signals, trying to see if I can train an Echo State Network, some experiments with eyecandy in SVG (the Design Doodles) and an Error Level Analysis tool I read about yesterday that can help determine whether a picture has been shooped or not ... and a whole random bunch of other little tryouts and experiments and doodads.
You?
Sounds like you have a good memory let me tell you that. I would love to see what you did, if you ever manage to get the hdd working again. (Maybe if you wrote an library for your synth?)
I hope you get back into it, demoscene can set off massive inspiration in others :) Do you have any other works?
Though I know how you feel about C++ ...
Damn I like the sound of your projects, especially the WiFi signal one.
I was ment to write a game for the DS (PsychoEx) but me and friend had falling out (Of course things are fixed now), ever since I started doing paid work for Programming I just can't find the motivation to write anything outside of work anymore ~ too lazy now. Just keeping it at PHP atm, working at Car Yard ... Its way too easy (and its consistant) - its keeping me fairly happy :D
PHP is easy to code, I agree, but it's not consistent IMO. It's lack of consistency is my biggest beef with it in fact :) Try any other loosely typed dynamic scripting language (ruby, python, perl ..) and you'll see what I mean. Take for example the array functions. About half of them start with array_ and the rest not. Some of them modify the array in place, while others return a new array, and some do both. Similar for the string functions, additionally half of them are named based on equivalent C functions, others have made-up names. The file functions ... etc :) In case you are interested, stuff is detailed here: http://www.tnx.nl/php.html though that page evangelizes Perl, which is IMO not really the best alternative (cause it's kind of old and therefore has its own fair share of weird quirks, newer languages like ruby and python are much better designed). In fact, really the only reason why PHP is still so popular is because it's a platform you can expect to be present at any random webhost everywhere, which is more a function of market share than the qualities of the language itself. Sorry it's a bit of a peeve of mine :)
Anyway. I didn't write a library for the synth, cause each instrument is just umm a few lines of C code :) Srsly once you know how, generating waveforms is not that hard. Kind of fun, like a modular synthesizer such as Buzz Machines, except even more freedom :)
I did write one VST Effect, if anybody is interested I can upload that. It's called Krush.dll and it does a kind of combo of randomized bitcrushing, lowpass filter and sinusoid waveshaping. I might just one day start writing VST effects again. Whenever I hear an autotuner, I actually want to code one, just to figure out if I can make them work, plus I got some novel ideas on how to do them which may sound more interesting than the done-to-death autotuners that are around these days.
Um, other work. Yes I made four 4K demos:
Design?! (http://www.pouet.net/prod.php?which=333), 2nd place Bizarre 1998
Organic (http://www.pouet.net/prod.php?which=332), 1st place Ambience 1999
Never Bored (http://www.pouet.net/prod.php?which=291), 2nd place TakeOver 1999
Meuk (http://www.pouet.net/prod.php?which=343), 2nd place Mekka & Symposium 2000
First one has no sound, second and third use the Adlib chip on your soundcard (real crappy hardware synth nobody uses anymore but is probably still on every soundcard) and the fourth actually uses a software synthesizer that generates the waveforms from various formulas, sawtooths, noise and filters. Unfortunately, because it tries to access the soundcard directly, very low-level, it's also the one that's least likely to run. I programmed all of them to run in DOS/Windows'98 and I don't think I ever got Meuk to run anymore even in Win2K. There is a tool called "DOSBox" I heard some people had success with, but I never got that working either. I did manage to rip just the soundtrack of Meuk, I can upload that later if you're interested.
If you can't find the motivation and want to start coding stuff for fun again, I may suggest learning Python. It really makes things so easy, you can write tiny fun experiments for whatever you can think of in just a few lines of code. XKCD is right on the money with this one: http://xkcd.com/353/
If entropy as is being discussed in this thread is related to the amount of predictability, then context is essential.
I have a recording of Roland Kirk playing live with Charles Mingus. It was the late 60s, so free jazz was in it's heyday. The guy is an absolute freak, and is throwing out atonal riffs, shrieks, and other abrasive and harsh tones.
However, this is to be expected.
The set ends with him blowing a low A on his sax. Surprisingly, this note goes on for about three minutes without stopping. Because he knows the technique as "circular breathing", he can hold a note far longer than most people would believe possible.
This is very unexpected.
"Jazz is the sound of surprise"
Quote from: Mangrove on April 22, 2010, 07:55:09 PM
"Jazz is the sound of fapping"
Fixed because I had to.
Quote from: Sigmatic on April 20, 2010, 09:12:20 PM
Who the hell would enjoy entropic, unpredictable music?
...anyone who likes Aphex Twin or Venetian Snares?
Or Ornette Coleman?
Quote from: Sigmatic on April 20, 2010, 11:08:40 PM
Alright, my bad. But please refrain from calling me "son", unless you happen to be at least twice my age.
So what's the point of studying entropy levels in music? Is it going to make better music, or is it just an exercise in measuring entropy?
It extends knowledge of psychology from reasons for visible aesthetic (the most pleasing form/number of shapes in a painting etc) to auditory aesthetic. There was a TED Talk about the former at some point.
Oh cool. Thanks for addressing that, I was a bit lost. :)
Yeah. Theres a certain amount of sensory complexity that a person will consider ideal; too little, and it's boring, too much and its noisy or overwhelming. This changes from person to person of course, but there are still trends over the general population.
Quote from: Kai on April 25, 2010, 07:56:11 PM
Yeah. Theres a certain amount of sensory complexity that a person will consider ideal; too little, and it's boring, too much and its noisy or overwhelming. This changes from person to person of course, but there are still trends over the general population.
And--I know, I am repeating myself--this is IMO extremely relevant to (my) Discordia, because it shows a Creative sweet spot of Novelty somewhere in the middle between Order and Disorder.
Quote from: Triple Zero on April 26, 2010, 08:00:45 AM
Quote from: Kai on April 25, 2010, 07:56:11 PM
Yeah. Theres a certain amount of sensory complexity that a person will consider ideal; too little, and it's boring, too much and its noisy or overwhelming. This changes from person to person of course, but there are still trends over the general population.
And--I know, I am repeating myself--this is IMO extremely relevant to (my) Discordia, because it shows a Creative sweet spot of Novelty somewhere in the middle between Order and Disorder.
Indeed.