[eu-gene] need advice on a framework design

[David Hart]

> But my comments were meant to be specific to the technique where the
> genotype is essentially mathematical expressions that are evaluated pixel
> by pixel to render a phenotype. I'm saying that specific genotype
> construction domain seems to have a limited aesthetic range.
>

I know what you meant, and largely I agree that historically it hasn't gone
very far.  But show me some generative art, of any kind, that has a wider
range.

The problem isn't the format of evolving math expressions that look nice
when plotted, the problem is we're choosing to building systems that
provide only indirect control, and we're baking aesthetic judgements into
those systems, whether we think so or not.  I assert that the, perhaps
overused, format of artificially evolving mathematical expressions is no
worse off than any other generative format we've seen thus far in terms of
aesthetic range.


There is, of course, a possible out. Just about any process could be
> referenced with a functional notation that can be treated like a
> mathematical expression in the genotype. So one could imagine calls to
> routines that fetch images from the web or send wireframes to 3D
> raytracers, etc. But at that point you're really outside of the domain I'm
> referring to.
>

So where's the line?  Your examples are at an extreme end, granted, but on
the other extreme end if we say the domain of evolved expressions is
strictly limited to a specific set of inputs that we deem "pure", and other
approaches are "outside of the domain", then of course you induce the very
problem you're concerned about.

If its considered "cheating", whatever that means, to put in anything more
than the simplest building blocks like gradients and sine functions and I'm
not allowed any combinators more complex than add or multiply, then I can
expect evolution to go very, very slowly.  If I can start applying some
larger constructs, or at the very least have the tools to build larger
constructs as I go, then I might be able to get somewhere.

To some degree, this notion that the building blocks of an evolutionary
system must be "pure" are the reason that the evolved images format is
stuck in a narrow range of output.  But its a false notion, noise and sine
functions are not the purest smallest possible mathematical building
blocks, and they already have an aesthetic value.  If we actually stuck to
pure building blocks, we'd be getting even less interesting results out of
image evolution.



> One reason the, in my view, exaggerated focus on the evolving expressions
> technique concerns me is that it has become an informal benchmark being
> used by some researchers working in the realm of computational aesthetic
> evaluation.
>

Maybe it makes a good benchmark for academic research precisely because it
is difficult to get very far away from the generators in the set?  That
will make it pretty obvious when the "evolution" is really doing its job
and the theory is working.



> My gut reaction is that even if an evaluation system could be developed
> that allowed human evaluation to be taken out of the evolutionary loop, it
> would probably turn out to be very specific to the peculiar evolving
> expressions look.  So the result wouldn't generalize and wouldn't really
> yield much leverage on or understanding of evolutionary art and design of
> other kinds. But again, that's just an intuition and not something I can
> demonstrably prove.
>

The theory claims otherwise, so if its impossible to escape a narrow range,
people using the technique should be asking why.   I acknowledge that its
*difficult* to escape the look, but I disagree that its impossible, and I
think I'm providing some small amounts of evidence to back up my claim with
some of my results, for example:

http://dahart.com/selection.jpg

I'm highlighting these particular samples for a very specific reason, not
because I like them.  They are examples of escaping the usual obvious
tells, and bunch of them demonstrate some mild success of the evolutionary
process by appearing to have shading and three dimensionality.  I selected
for those attributes, it is not some inherent part of the building blocks I
used.  I'm using the same building blocks everyone else is, and sometimes
selecting for shape and shading, and sometimes achieving it.  This suggests
the evolutionary theory is working, as well as suggesting that it is in
fact possible to escape the well known look, even if I didn't get all that
far.

--
David.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.generative.net/pipermail/eu-gene/attachments/20120124/ed7e34a0/attachment-0001.htm