Seeking examples of human-competitive results produced by EC

[John Koza <koza@smi.stanford.edu>]

Hello:

I've believed for some time that the fields of artificial intelligence and
machine learning have become trapped on the local optimum of solving and
re-solving academically created "toy" problems.  I've also believed that one
of the strong points for the field of evolutionary computation is that it
has produced numerous examples of non-trivial results that are competitive
with human-produced results.

I'm writing an article that will include a list of these human-produced
results from the EC field.  I'd like the list to be as complete and current
as possible and would like to know about any examples that I've missed on my
FIRST ROUGH DRAFT of a list.

I'm particularly interested in examples that satisfy any one of the eight
criteria on the list below (although I'd also be equally interested in any
that satisfy any other similarly stringent arms-length criteria).

THE 8 CRITERIA:
An automatically created solution to a problem is competitive with
human-produced results if it satisfies any of the eight criteria below (or
any other similarly stringent and arms-length criterion).

Note that each of the criteria are couched in terms that are entirely
external to the fields of machine learning, artificial intelligence, and EC.

Note that the criteria are RESULTS-oriented, not METHOD-oriented.  That is,
each criteria requires the RESULT that stands on its own merit --- not on
the fact that the result was mechanically produced.

Note also that the criteria all require a direct test that can be performed
and verified by an outsider.  This excludes, as I think necessary for a list
like this, results of the type of "the EC result is 25% better than what our
company was previously doing, but sorry I can't tell you what we were
previosly doing or doing now."

(A) The result was patented as an invention in the past, is an improvement
over a patented invention, or would qualify today as a patentable new
invention.

(B) The result is equal to or better than a result that was accepted as a
new scientific result at the time when it was published in a peer-reviewed
scientific journal.

(C) The result is equal to or better than a result that was placed into a
database or archive of results maintained by an internationally recognized
panel of scientific experts.

(D) The result is publishable in its own right as a new scientific
result --- independent of the fact that the result was mechanically created.

(E) The result is equal to or better than the most recent human-created
solution to a long-standing problem for which there has been a succession of
increasingly better human-created solutions.

(F) The result is equal to or better than a result that was considered an
achievement in its field at the time it was first discovered.

(G) The result solves a problem of indisputable difficulty in its field.

(H) The result holds its own or wins a regulated competition involving human
contestants (in the form of either live human players or human-written
computer programs).

------------------

Here are examples that I've located:

--- Creation of a sorting network that is superior to the previously known
human-created solution to the same problem (Juille 1995)

--- creation of a cellular automata rule that outperformed previous
human-written algorithms for the same problem (Juille and Pollack 1998)

---  design for the shape of an aircraft wing created by the genetic
algorithm that was patented (Kroo, McMasters, and Pavek 1995)

--- design of a wire antenna that was patented (Altshuler and Linden 1998)

--- evolution using EP/ES of a checker player that plays as well as human
class B checkers (Fogel and Chellapilla 1999).



John R. Koza

Consulting Professor
Stanford Medical Informatics
Department of Medicine
Medical School Office Building
Stanford University
Stanford, California 94305

Consulting Professor
Department of Electrical Engineering
School of Engineering
Stanford University

Phone: 650-941-0336
Fax: 650-941-9430
E-Mail: koza@stanford.edu
WWW Home Page: http://www.smi.stanford.edu/people/koza

For information about field of genetic programming:
http://www.genetic-programming.org

For information about Genetic Programming Inc.:
http://www.genetic-programming.com

For information about GECCO-2000 (GP-2000) conference in Las Vegas on July
8 -12, 2000, visit:
http://www.genetic-algorithm.org/GECCO2000/gecco2000mainpage.htm