Re: Gene Expression Programming(GEP)

h2s schrieb:

Thanks also for the information.do you object to her naming it GEP or
to the methodology and algorithm she use?am still a newcomer as you
said.

In my opinion it’s okay that she names it like this, although I see that
it could confuse people. But everyone who works in this field will
sooner or later learn about the differences.
GEP works, no question. It just is GP with some extra step.
It converts a tree (if doing GP in Lisp then tree means a program in
that case) into another representation on which the genetic operations
are then performed. To test the programs they have to be converted back
into tree form and then get evaluated (in GEP).
My only point is that I see not why the claim that GEP works 100-60k
times “better” than GP, only because the genetic operations are not done
on the tree itself.
We could as well use WinRAR to pack a lisp program, then shuffle the
letters in the .rar file around and unpack it back to a Lisp program and
expect it to evolve dramatically much faster as it would if you manipulate
the tree directly, as it happens in GP.
Okay, I admit that .rar files are a bad example here because one couldn’t
unpack a randomly changed file back to a working Lisp program, but you
got the idea.

I personally suggest to at least think about getting the books of John
Koza who is one of the leading experts in the world in the area of
Genetic Programming (not really the inventor, but most people regard
him more or less like that). Plus he writes his books in a style so that
even mortals can understand him. He does not try to seem extremly clever
by using complicated descriptions that only a few people could decipher.
Very good job in writing books.
One just needs a background in thinking and it’ll be good ;-)


André
--
.

Relevant Pages

  • Re: 6502 FPGA core
    ... Lisp program would be to use some form of syntax trees. ... Producing VHDL code from the tree representation should be ...
    (comp.lang.lisp)
  • Re: "a language of failed dreams"
    ... > clocking 1489 reviews (which might change in the time it takes me to ... > which would give him an incredible capacity of almost two books a day. ... Does anyone have a Lisp program to make anagrams of his name? ...
    (comp.lang.lisp)