From: Jeff Bone (jbone@jump.net)
Date: Thu Apr 06 2000 - 19:08:45 PDT
> I could also catch a word with Koza (Beowulf list residents
> know he's running a 1 kNode (half a kilonode, actually, considered
> that he is using dual-CPU boxes) Beowulf dedicated to GP (since
> recently, Java). He's aware of bad shape of the the fitness landscape
> when it comes to shotgunning the machine code bits (no fitness-neutral
> filaments permeating the high-dimensional machine opcode space),
I'm very surprised that Koza is looking at this problem from this angle, if
that's in fact what your comment indicates. Isn't GP different from the
"shotgunning the bits" approach of GA? IIRC, it's based on mutating a
parsetree-like higher-order representation of a computation, and is therefore
more abstractly about breeding algorithms (for which the phase space has less
dimensions?) (At least, this is what I recall from Koza's GP tome. It was
all about twiddling around lisp-like representations of algorithms, wasn't
it?)
>
> If I was trying to make the Net transcend, I'd figure out a way to
> mutate machine code robustly (using ALife on mutation functions), and
> applied that to distributed neural net simulators plus remote exploit
> finders. Of course, creating easily machine-readable content makes the
> task easier, once the golem awakens.
>
Nah, the if you just twiddle bits you can do it more cheaply but you probably
get more crap than if you mutate some higher-order, more abstract
representation. I imagine the pure-GP approach has better cost / benefit
characteristics than brute GA. (NB: those terms should actually be
reversed... GP is more about algorithms, while GA seems to be more about
programming at the bitstream level.)
>
> > Oh, and Carey: very good of you to take notes, and thank you
> > for posting them here! May i forward them to others who were
> > not able to attend?
>
> Indeed, it was a valuable service. I wish I had taken time to take notes.
No kidding. Huge kudos, especially from my last-minute-pullout-self. ;-)
jb
This archive was generated by hypermail 2b29 : Thu Apr 06 2000 - 19:14:38 PDT