Thinksamuel

As a student of composition, I always felt a lack of a method for algorithmic composition. For counterpoint and harmony, there are treatises and structured methods but for the 20th century idiom with all its facets, there is no structured pedagogical means for students to learn about algorithmic composition. I would like to write a method book to allow students to learn about algorithmic composition, giving them information and i the end of each chapter some exercises. It would be like the book of Straus about posttonal theory, only it will focus on algorithmic methods.

I have spoken to my professor about this but he told me I need to know how algorithmic composition is taught in other schools. For this I would like to get to know from you: How did you learn algorithmic composition? If yo are a teacher, how do you teach algorithmic composition to your students?

I am hoping to get many answers to enhance my research proposal as best as possible.

As a student of composition, I always felt a lack of a method for algorithmic composition. For counterpoint and harmony, there are treatises and structured methods but for the 20th century idiom with all its facets, there is no structured pedagogical means for students to learn about algorithmic composition. I would like to write a method book to allow students to learn about algorithmic composition, giving them information and i the end of each chapter some exercises. It would be like the book of Straus about posttonal theory, only it will focus on algorithmic methods. 

how-did-you-learn-algorithm-composition

the website in that forum link is great. a fantastic tool and learning curve, i go on every now and again, see what is new. 

the website in that forum link is great. a fantastic tool and learning curve, i go on every now and again, see what is new. 
i would highly recommend it as well...


there is math, and there is improvisation. algorithmic 

you can learn math by using maxmsp, and you can find out about

the possibilties of using math to compose by simply trying it 

the majority of possible algorithms is completly useless for 

music, yet music always contains lot of math.

use some basic philosophical ideas to start composing training

sessions: compose a piece of music backwards, use an indian

scale, do a 11/8 signature song or compose a piece which only

consists of click sounds. such exercises will bring up new ideas

on the "event" or "math" level of your work so that you do not

fall back in the try-and-listen-how-it-feels improvisation scheme.

there is math, and there is improvisation. algorithmic 
composition is somewhere in between.

you can learn math by using maxmsp, and you can find out about
the possibilties of using math to compose by simply trying it 
out.

the majority of possible algorithms is completly useless for 
music, yet music always contains lot of math.

use some basic philosophical ideas to start composing training
sessions: compose a piece of music backwards, use an indian
scale, do a 11/8 signature song or compose a piece which only
consists of click sounds. such exercises will bring up new ideas
on the "event" or "math" level of your work so that you do not
fall back in the try-and-listen-how-it-feels improvisation scheme.

How'd I learn? I Read Xenakis. Read Hiller. Read Koenig. Read the relevant chapters of Dodge/Jerse and Roads' 

. Read some of Cope. Read swathes of Barlow. Also read Boulez, Stockhausen, Messiaen, Perle, Stravinsky, Forte, Hindemith, Cowell, Cage, Lewin and lots of other texts that weren't directly related to AC but with insights into compositional techniques. These have all helped make my approach to AC what it is. (Please do not hold it against these people;-)

At some point I started doing it as well as talking about it with other composers who were interested.

I don't know that many schools teach AC as a curriculum. But there are a number of texts extant (see above) that discuss topics in AC to varying degrees of competence. There's also some Temperley on my bookshelf waiting to be read, and I've got to get around to Gareth Loy's books. You may also want to look at Robert Rowe's and Todd Winkler's books. Review Computer Music Journals of the last 25 years.

There is also some dead wood out there. I'd suggest you start with some of the above, then go hunting.

How'd I learn? I Read Xenakis. Read Hiller. Read Koenig. Read the relevant chapters of Dodge/Jerse and Roads' Tutorial. Read some of Cope. Read swathes of Barlow. Also read Boulez, Stockhausen, Messiaen, Perle, Stravinsky, Forte, Hindemith, Cowell, Cage, Lewin and lots of other texts that weren't directly related to AC but with insights into compositional techniques. These have all helped make my approach to AC what it is. (Please do not hold it against these people;-)

As someone who is both a programmer and composer, I feel

like I am constantly trying unite these two worlds. The

thought that I could take a powerful tool like the

computer and have it help me unlock something more meaningful,

I think it is very important to keep in mind the environment

you are working in. Having an environment that allows you

to examine the behavior of an algorithm is crucial to figuring

out if it might be useful to you. That may mean you need to

visualize the output, or maybe hear the output, Max is great for

Get to know a language. Max supports Python, Java, C++, LISP,

all have different strengths and weaknesses. But the more you

know a language the easier it is get ideas in your head, out

in the real world, where you can experiment with then and

Don't reinvent the wheel. If someone has developed a tool

that does something similar give it a try. It will give you

great insight as to how useful something is and if it is

worth the time to re-implement it yourself.

It is all about mapping. An important question to always

ask yourself is: how can you take the behavior

of an algorithm and map it to musical parameters so that

that it creates a certain effect or creates a sense of change

in the structure of the music. It takes a lot of experimentation.

Sometimes with little result, sometimes with spectacular result.

You have to be very patient with yourself. Allow yourself the

freedom to go off on a tangent even though you may at first

think it is a frivolous exercise. That is how great discoveries

Ultimately I believe the most effective use of algorithmic

composition is to employ it to create a desired objective.

Designing an algorithm that makes musical decisions for you

never sounds that interesting. You can employ probability to

add variety and diversity, but ultimately you are the composer,

you need to have a higher level view of how an algorithm fits

and contributes to the whole of the piece.

Also I would recommend the book "Notes from the Metalevel"

by Heinrick Taube. It is a great introduction to algorithimic 

composition techniques. His algorithmic comp environment is 

called Common Music, which uses LISP. It also supports a simpler 

scripting language called SAL. Definitely check it out.

As someone who is both a programmer and composer, I feel
like I am constantly trying unite these two worlds. The
thought that I could take a powerful tool like the
computer and have it help me unlock something more meaningful,
is a very compelling drive.

I think it is very important to keep in mind the environment
you are working in. Having an environment that allows you
to examine the behavior of an algorithm is crucial to figuring
out if it might be useful to you. That may mean you need to
visualize the output, or maybe hear the output, Max is great for
this.

Get to know a language. Max supports Python, Java, C++, LISP,
all have different strengths and weaknesses. But the more you
know a language the easier it is get ideas in your head, out
in the real world, where you can experiment with then and
hear them.

Don't reinvent the wheel. If someone has developed a tool
that does something similar give it a try. It will give you
great insight as to how useful something is and if it is
worth the time to re-implement it yourself.

It is all about mapping. An important question to always
ask yourself is: how can you take the behavior
of an algorithm and map it to musical parameters so that
that it creates a certain effect or creates a sense of change
in the structure of the music. It takes a lot of experimentation.
Sometimes with little result, sometimes with spectacular result.
You have to be very patient with yourself. Allow yourself the
freedom to go off on a tangent even though you may at first
think it is a frivolous exercise. That is how great discoveries
are made.

Ultimately I believe the most effective use of algorithmic
composition is to employ it to create a desired objective.
Designing an algorithm that makes musical decisions for you
never sounds that interesting. You can employ probability to
add variety and diversity, but ultimately you are the composer,
you need to have a higher level view of how an algorithm fits
and contributes to the whole of the piece.

Also I would recommend the book "Notes from the Metalevel"
by Heinrick Taube. It is a great introduction to algorithimic 
composition techniques. His algorithmic comp environment is 
called Common Music, which uses LISP. It also supports a simpler 
scripting language called SAL. Definitely check it out.


Algorithmic composition is a great way to learn about meta levels. In a way composing and using algorithms is a contradiction, if you only look at one level of perspective.

The only compositional skill you'd really need is maybe counterpoint. 

But in a very general way, not as a collection of rules, but of an ability of analytical listening. Studying other composers, not for copying what they did, but to get a feeling for how they shaped their works.

Then you need some other skills, but those you will develop by doing. Getting into programming Max or Supercollider or csound or Lisp or Pascal (my own jump into it) is a good start obviously...

Algorithmic composing is composing rules instead of composing single events (old fashioned notes for example). But still you need to compose events on a higher level. For example changing of rules, and if you create a rule for changing rules you still need to find out how to begin and how to end. Without composing these meta events as events, you would end up creating sounds instead of music. Creating complex sounds is actually a very obvious result of this path and can be easily used in wonderful sound installations.

The picture I had when starting, was the sound of a burbling creek. It is always the same, and never the same. It is never boring, you can listen to it for ages. I only stop tweaking my algorithms if it has this quality, being able to listen to it with fun for hours...

Algorithmic thinking on the other hand is a good method to focus on the musical intention of a piece. 

For example if you want to focus on the pitched sound of a cello, how to lower the attention to musical parameters which have the natural focus in traditional music? Usually this is melody, harmony, rhythm etc. If you just avoid melody by using twelve tone rows, you still would have the focus on it, as it is opposing old ways of doing it. If you play only a single pitch or avoid pitched sounds, you limit the sounds you want to explore. 

One common solution is already the most simple algorithm: repetition as exaggerated in minimal music, which can lead the audience into a trance like state which might be something you also don't want...;-).

With algorithmic composition you can turn any musical parameter into a sound by simply applying the same rule and not altering it through out the piece... This will make events that don't follow rules stand out...

And please don't do it because a composer has to have an algorithmic piece, a string quartet and a symphony in his repertoire to be accepted as a serious composer. This is only something for boring careerists who believe the label on their card and a degree is the only thing it needs to become a composer...

Learning algorithmic composition is not possible by reading a book, as you don't learn composition by reading books about counterpoint and harmony. You need to get your hands dirty and need colleagues to discuss. A book can help to get started though, but explain right in the beginning that none of the examples in the exercises is allowed to be used in their own compositions...;-)

The only compositional skill you'd really need is maybe counterpoint. 
But in a very general way, not as a collection of rules, but of an ability of analytical listening. Studying other composers, not for copying what they did, but to get a feeling for how they shaped their works.

Algorithmic thinking on the other hand is a good method to focus on the musical intention of a piece. 
For example if you want to focus on the pitched sound of a cello, how to lower the attention to musical parameters which have the natural focus in traditional music? Usually this is melody, harmony, rhythm etc. If you just avoid melody by using twelve tone rows, you still would have the focus on it, as it is opposing old ways of doing it. If you play only a single pitch or avoid pitched sounds, you limit the sounds you want to explore. 
One common solution is already the most simple algorithm: repetition as exaggerated in minimal music, which can lead the audience into a trance like state which might be something you also don't want...;-). 

Learning algorithmic composition is not possible by reading a book, as you don't learn composition by reading books about counterpoint and harmony. You need to get your hands dirty and need colleagues to discuss. A book can help to get started though, but explain right in the beginning that none of the examples in the exercises is allowed to be used in their own compositions...;-)


stefantiedje wrote on Wed, 18 November 2009 07:40

One thing I might add. Algorithmic composition need not be rule 

based. Most of the Algcomp I do is computational geometry

mapped on to musical parameters. It is higher level

forms projected on to musical spaces (i.e. pitch, timbre,

rhythm, etc.) This allows me the ability to control the 

shapes of phrases and gestures (at both the micro and macro

level) while maintaining precise output of form.

stefantiedje wrote on Wed, 18 November 2009 07:40
Algorithmic composing is composing rules instead of composing single events (old fashioned notes for example). But still you need to compose events on a higher level. For example changing of rules, and if you create a rule for changing rules you still need to find out how to begin and how to end. Without composing these meta events as events, you would end up creating sounds instead of music. Creating complex sounds is actually a very obvious result of this path and can be easily used in wonderful sound installations.

One thing I might add. Algorithmic composition need not be rule 
based. Most of the Algcomp I do is computational geometry
mapped on to musical parameters. It is higher level
forms projected on to musical spaces (i.e. pitch, timbre,
rhythm, etc.) This allows me the ability to control the 
shapes of phrases and gestures (at both the micro and macro
level) while maintaining precise output of form. 

How did you learn algorithm composition?