Genetic programming: on the programming of computers by means of natural selection | Guide books

Genetic programming: on the programming of computers by means of natural selectionDecember 1992

Author:
John R. Koza
Stanford Univ., Stanford, CA

Publisher:

MIT Press
55 Hayward St.
Cambridge
MA
United States

ISBN:978-0-262-11170-6

Published:21 December 1992

Pages:

680

Available at Amazon

Bibliometrics

Sections

1992

Abstract

No abstract available.

Cited By

Contributors

John Reed Koza
Stanford University
- Publication Years1972 - 2010
- Publication counts72
- Citation count3,945
- Available for Download19
- Downloads (cumulative)6,431
- Downloads (12 months)494
- Downloads (6 weeks)66
- Average Downloads per Article338
- Average Citation per Article55
View Full Profile

Index Terms

Genetic programming: on the programming of computers by means of natural selection

Reviews

Reviewer: Louis Hodes

Genetic programming (book) One main idea—the generation of programs by massive trial and error—is developed in this book. Koza begins with a set of random programs in generation zero. He uses the analogy of biological evolution, so the most successful programs have an increased chance to reproduce and recombine in each succeeding generation. The author uses a standard format whereby the programs consist of a set of inputs and a set of functions capable of being applied hierarchically in a LISP-like composition. The inputs and functions vary according to the specific application, but the programs can always be viewed as tree structures. This structure allows genetic recombination to exchange intact subtrees. Thus, meaningful offspring are likely. The above method is applied to dozens of problems in a wide variety of areas. None of the examples realizes a useful way to perform a useful task, however. The difficulty lies in the difference between genetic programming and natural selection as it occurs in nature. Koza's examples all need explicit detailed fitness measures, which often are already program solutions. In any case, determining fitness for all the programs of every generation is the most time-consuming and limiting factor of this approach. The first four chapters are introductory, giving background material on machine learning and genetic algorithms. Chapter 5 describes genetic programming. The details are left for chapter 6, which covers such questions as how the fitness measure is normalized. Chapters 7 through 9 give some examples and compute statistics on population sizes for some tasks. Chapters 10 to 23 each introduce a new form of application, with some detailed examples. Chapter 10 covers error-driven evolution. Chapter 11 discusses cost-driven evolution. Chapter 12 discusses behavior, using the example of artificial ant feeding. Chapter 13 addresses subsumption, with the example of a robot finding and pushing a box. Chapter 14 considers entropy, including random number generation. Chapter 15 is about strategy and considers the case of minimax games. Chapter 16, on coevolution, studies both-sides-minimax games. Chapter 17 covers classification, including distinguishing two spirals. Chapter 18 describes iteration, using sequence finding as an example. Chapter 19 presents constraint syntax. Chapter 20 covers building blocks. Chapter 21 discusses hierarchy, with the example of parity functions. Chapter 22 addresses parallelization. Chapter 23, on ruggedness, uses the example of a biathlon. The biathlon example alternates the fitness measure so the same program learns two endpoints—symbolic regression and artificial ant feeding, depending on interpretation . In the remaining chapters, the author discusses extraneous inputs and functions, considers some other issues, and compares genetic programming with other artificial intelligence paradigms. A long chapter is devoted to spontaneous generation of self-replicating programs. The implementation details are presented in a set of appendices occupying 88 pages. The purpose of this book is to convince the reader of the importance of genetic programming by the multitude and variety of applications. It fails because none of the examples rises to the level of actual utility. We have reached the stage where we expect our artificial intelligence products to be competitive. The best feature of this work is the structure-preserving crossover resulting from the tree structure of the programs. We need to look for such mechanisms to contain meaningful blocks in biological recombination. Genetic programming (videotape) The one-hour video covers the same ground as the book without much of the background and details. It is especially appropriate in presenting the tree structure of the programs with crossover as an exchange of subtrees. About 20 of the examples from the book are covered. Many of these examples, such as artificial ant feeding, provide entertaining animation. The function hierarchy capability applied to the 11-parity function is explained well. The videotape provides a fast, painless way to learn genetic programming. I recommend it to those who remain curious about genetic programming after reading the above book review.

Access critical reviews of Computing literature here

Become a reviewer for Computing Reviews.

Comments

Please enable JavaScript to view thecomments powered by Disqus.

Recommendations

Using genetic programming for the induction of novice procedural programming solution algorithms
SAC '02: Proceedings of the 2002 ACM symposium on Applied computing

This paper describes a genetic programming system for the induction of solutions to novice procedural programming problems. This genetic programming system will form part of a generic architecture for the development of intelligent programming tutors ...
Genetic programming: a tutorial introduction
GECCO '12: Proceedings of the 14th annual conference companion on Genetic and evolutionary computation

Genetic programming emerged in the early 1990's as one of the most exciting new evolutionary algorithm paradigms. It has rapidly grown into a thriving area of research and application. While sharing the evolutionary inspired algorithm principles of a ...
Genetic Programming: A Tutorial Introduction
GECCO '16 Companion: Proceedings of the 2016 on Genetic and Evolutionary Computation Conference Companion

Genetic programming emerged in the early 1990's as one of the most exciting new evolutionary algorithm paradigms. It has rapidly grown into a thriving area of research and application. While sharing the evolutionary inspired algorithm principles of a ...

Save to Binder