Genetic Programming IV: Routine Human-Competitive Machine Intelligence
Created by W.Langdon from
gp-bibliography.bib Revision:1.2031
@Book{koza:gp4,
author = "John R. Koza and Martin A. Keane and
Matthew J. Streeter and William Mydlowec and Jessen Yu and
Guido Lanza",
title = "Genetic Programming {IV}: Routine Human-Competitive
Machine Intelligence",
publisher = "Kluwer Academic Publishers",
year = "2003",
keywords = "genetic algorithms, genetic programming",
ISBN = "1-4020-7446-8",
URL = "
http://www.genetic-programming.org/gpbook4toc.html",
abstract = "Genetic programming (GP) is method for automatically
creating computer programs. It starts from a high-level
statement of what needs to be done and uses the
Darwinian principle of natural selection to breed a
population of improving programs over many
generations.
Genetic Programming IV: Routine Human-Competitive
Machine Intelligence presents the application of GP to
a wide variety of problems involving automated
synthesis of controllers, circuits, antennas, genetic
networks, and metabolic pathways.
The books describes 15 instances where GP has created
an entity that either infringes or duplicates the
functionality of a previously patented 20th-century
invention, 6 instances where it has done the same with
respect to post-2000 patented inventions, 2 instances
where GP has created a patentable new invention, and 13
other human-competitive results.
The book additionally establishes:
GP now delivers routine human-competitive machine
intelligence.
GP is an automated invention machine.
GP can create general solutions to problems in the form
of parameterised topologies.
GP has delivered qualitatively more substantial results
in synchrony with the relentless iteration of Moore's
Law.",
notes = "PDF of chapter 1 available. 42-minute DVD included in
Jaws-4
-------------------- Comments on the Book
-------------------- The research reported in this book
is a tour de force. For the first time since the idea
was bandied about in the 1940s and the early 1950s, we
have a set of examples of human-competitive automatic
programming. John H. Holland, University of Michigan
In 1992, John Koza published his first book on genetic
programming and forever changed the world of
computation. At the time, many researchers, myself
included, were skeptical about whether the idea of
using genetic algorithms directly to evolve programs
would ever amount to much. But scores of conquered
problems and three additional books makes the case
utterly persuasive. The latest contribution, Genetic
Programming IV: Routine Human-Competitive Machine
Intelligence, demonstrates the everyday solution of
such holy grail problems as the automatic synthesis of
analog circuits, the design of automatic controllers,
and the automated programming of computers. This would
be impressive enough, but the book also shows how to
evolve whole families of solutions to entire classes of
problems in a single run. Such parametric GP is a
significant achievement, and I believe it foreshadows
generalised evolution of complex contingencies as an
everyday matter. To artificial evolutionaries of all
stripes, I recommend that you read this book and breath
in its thoughtful mechanism and careful empirical
method. To specialists in any of the fields covered by
this books sample problem areas, I say read this book
and discover the computer-augmented inventions that are
your destiny. To remaining skeptics who doubt the
inventive competence of genetics and evolution, I say
read this book and change your mind or risk the strong
possibility that your doubts will soon cause you
significant intellectual embarrassment. David E.
Goldberg, University of Illinois
The adaptive filters and neural networks that I have
worked with over many years are self-optimising systems
where the relationship between performance (usually
mean-square-error) and parameter settings (weights) is
continuous. Optimization by gradient methods works well
for these systems. Now, this book describes a wider
class of optimisation problems where the relationship
between performance (fitness) and parameters is highly
disjoint, and self-optimization is achieved by
nature-inspired genetic algorithms involving random
search (mutation) and crossover (sexual reproduction).
John Koza and his colleagues have done remarkable work
in advancing the development of genetic programming and
applying this to practical problems such as electric
circuit design and control system design. What is
ingenious about their work is that they have found ways
to approach design problems by parameterizing both
physical and topological variables into a common code
that can be subjected to genetic programming for
optimisation. It is amazing how this approach finds
optimised solutions that are not obvious to the best
human experts. This fine book gives an accounting of
the latest work in genetic programming, and it is must
reading for those interested in adaptive and learning
systems, neural networks, fuzzy systems, artificial
intelligence, and neurobiology. I strongly recommend
it. Bernard Widrow, Electrical Engineering Department,
Stanford University
John Koza's genetic programming approach to machine
discovery can invent solutions to more complex
specifications than any other I have seen. John
McCarthy, Computer Science Department, Stanford
University
",
size = "pages",
}
Genetic Programming entries for
John Koza
Martin A Keane
Matthew J Streeter
William J Mydlowec
Jessen Yu
Guido Lanza