Matt Gough Blog [closed]: fractals? or other
Tuesday, May 11 2004, 01:15
when visiting michael and davide the subject of fractal choreography was raised. there wasn't the chance to talk about this more in depth but I has been on my mind ever since. I have always been uncomfortable with claims of fractal choreography [1,2] but i'm only just getting round to document my thoughts and reasoning. rather than referring to our conversion I look at the two posts they have made on choreograph.net that can be found here.
I suggest you read both posts and the word document first. I should also point out that the posts are dated 2002 [michael] and 2003 [davide] and therefore probably do not represent their current thinking.
michael and davide, I think that there is a difference between what you are talking about and the examples that you are giving. rather than fractals you have described Genetic Algorithms [GA], although they may be 'fractal like' they are not fractals;
fractals are not 'self-similar patterns whose parts bear a resemblance on different scales', fractals are self-similar on all scales
davide's experiment does not manifest this property [the recognisable patterns are linear and not scalably self similar] and michael's structure [whilst more fractal in nature] would create a mass of repeated movement that would be exceptionally difficult to perform down to a gestural level but technically would need to be 'sub' gestural, whilst expanding to the performance space itself. two personal points;
fractal movement can not use whole body postures as a seed unless the 'limb trees' could replicate the global pose and transformations. davide's experiment with Nonlinear Generators is a good example of a GA or evolutionary algorithm rather than a fractal.
the GA can be observed in the treatment of the 'seed population' of ten movements he gave the dancers [the first generation pool]. the second third and 'n' generations are all generated with the same operators;
due to human cognitive and physical nature this 'list' evolution will [unconsciously] contain crossover and mutation operators [re mapping of other dancers movements and adaption of existing movements].
davide's experiment also fulfills the the basic aspect of genetic algorithms design:
as you can see it is genetic evolution that becomes 'complex through the interplay of variables' with and endless number of variations, not the implementation of fractals. fractal set display particular properties despite variances in variables [see here and here]. althougth this is also true for GA's, Iterated Function Systems [IFS] and other fractal like systems.
personally I think that implementing 'true' fractals as a choreographic structure would be impossible unless a particular part of the dancer was tracing and imaginary or real line that formed a fractal shape. the human body is not pliable enough, and the brain not ordered enough to represent fractal in or with the whole body of a single dance. it would be possible to have a large scale relatively static fractal with many dancers.
GA's however are more easier to implement as davide has shown. there are many different operator that one can choose from and apply to trees, lists or arrays to create movement or structures. there is also the option to use GA's to make dances 'on the fly' or to use GA's to create set scores which the dancer performs. both options are valid and have differing uses and applications.
I am starting work or a dance based on Lindenmayer systems [l-systems] whose recursive, self similar, fractal like nature can be used to describe plants and complex branching structures such as the nervous system.
various l-system based examples can be created, some that you may have come across; fibonacci numbers, cantor dust, a variant of the Koch snowflake [only using right angles] and more interestingly [for 'realtime' movement generation] median space filling curves such as Levy's dragon curve [non intersecting curves which can be iterated to produce more and more complexity]
rather than setting out only to use realtime creation I have decided to created movement scores based on the results of non deterministic and context sensitive l-system GA's. this re interpretation of the results [graphical representations] leads to less fractal like systems that although having self similar properties are somewhat more complex and abstract. this approach also is suited to experimenting within the xdsn notation system. in addition to the set scores i will also use realtime GA's for movement generation.
this is the type of image i am working with for the fixed score. this is a small gif file for quick preview, it does not show much detail but it gives the idea.
you can download the svg of this image if you have an svg viewer or want to examine the code in a text editor
yet notwithstanding my own approach I would love to see further investigation into realtime implementations of GA's. I can't think of anyone who is doing this. I don't know if davide still has an interest in this but i'm sure that some interesting work could result.
rather than being just a choreographic tool the use of math in this particular apllication is and site and context specific.
[1]
A good day's work. Morning with Jill on some 'fractal choreography' - building up simple rule sets which create a self-evolving, complex dance that never repeats itself http://creations.morle.com/portfolio/performance/shift/[2]
In "The Gism from the schism of the isms: hey but where's the egg, Mama? (in truncated form)," Suber juxtaposes the aesthetics of movement with the quantitative theory of mathematical indeterminacy. While none of the movement in "Gism" is predetermined, it is not impro visation. There is order to this chaos. Together with the dancers, Suber created a complicated sequence of movement that included 412 different steps and eleven phrases, as well as directional, sequential and orientational twists. While Suber out lined this concept of the fractal dance, it is up to the dancers to meld the fragments creatively into aesthetically pleasing, fluid movements, to balance the complexity and order inherent in the process."It is like a jigsaw puzzle," explains Suber. "We mix up the phrases, steps, sequence and direction, and then we try to put it back together. I like to think of it as graceful twister." In addition, Suber has been writing text that reflects the same indeterminacy of the movement, lighting and sound.
http://www.news.cornell.edu/Chronicle/96/3.7.96/dance96.html
[*]as a note of interest some areas of robotics are using GA's in higher functioning robots [20 degrees of freedom upwards] to automate limb functions.
Filed by Article19 at 1:15 PM | 
Stumble It!