The Troubling Aspects of a Building Block Hypothesis
for Genetic Programming
U. O'Reilly, и F. Oppacher. Foundations of Genetic Algorithms 3, стр. 73--88. Estes Park, Colorado, USA, Morgan Kaufmann, (31 July--2 August 1994)Published 1995.
Аннотация
In this paper we carefully formulate a Schema Theorem
for Genetic Programming (GP) using a schema definition
that accounts for the variable length and the
non-homologous nature of GP's representation. In a
manner similar to early GA research, we use
interpretations of our GP Schema Theorem to obtain a GP
Building Block definition and to state a ``classical''
Building Block Hypothesis (BBH): that GP searches by
hierarchically combining building blocks. We report
that this approach is not convincing for several
reasons: it is difficult to find support for the
promotion and combination of building blocks solely by
rigourous interpretation of a GP Schema Theorem; even
if there were such support for a BBH, it is empirically
questionable whether building blocks always exist
because partial solutions of consistently above average
fitness and resilience to disruption are not assured;
also, a BBH constitutes a narrow and imprecise account
of GP search behaviour.
FOGA-3 An early version of this paper is an SFI
Technical Report (see OReilly:1992:tabbGP). The
paper was accepted to a conference called ``Foundations
of Genetic Algorithms III''. It was presented at the
conference in Estes Park, Co, in July of 1993 (1994?)
and the latest version will appear in proceedings
forthcoming this year.
Presents a schema theorem for genetic programming based
on analogy with Holland's ST. Definition of schema more
general than Koza's (in Koza:book). Also delvelops
building block hyposthesis (again using analogy with
linear GA's). Then comprehensively trashes them both!
Some arguments against them apply equally well to
linear (bit string) GAs.
Chapter 4 of U.M. O'Reilly's PhD thesis
OReilly:thesis is similar to this paper. Pages
192-196 of Chapter 7 summarize the results.
%0 Conference Paper
%1 OReilly:1995:tabbGP
%A O'Reilly, Una-May
%A Oppacher, Franz
%B Foundations of Genetic Algorithms 3
%C Estes Park, Colorado, USA
%D 1994
%E Whitley, L. Darrell
%E Vose, Michael D.
%I Morgan Kaufmann
%K algorithms, genetic programming
%P 73--88
%T The Troubling Aspects of a Building Block Hypothesis
for Genetic Programming
%U http://citeseer.ist.psu.edu/oreilly92troubling.html
%X In this paper we carefully formulate a Schema Theorem
for Genetic Programming (GP) using a schema definition
that accounts for the variable length and the
non-homologous nature of GP's representation. In a
manner similar to early GA research, we use
interpretations of our GP Schema Theorem to obtain a GP
Building Block definition and to state a ``classical''
Building Block Hypothesis (BBH): that GP searches by
hierarchically combining building blocks. We report
that this approach is not convincing for several
reasons: it is difficult to find support for the
promotion and combination of building blocks solely by
rigourous interpretation of a GP Schema Theorem; even
if there were such support for a BBH, it is empirically
questionable whether building blocks always exist
because partial solutions of consistently above average
fitness and resilience to disruption are not assured;
also, a BBH constitutes a narrow and imprecise account
of GP search behaviour.
%@ 1-55860-356-5
@inproceedings{OReilly:1995:tabbGP,
abstract = {In this paper we carefully formulate a Schema Theorem
for Genetic Programming (GP) using a schema definition
that accounts for the variable length and the
non-homologous nature of GP's representation. In a
manner similar to early GA research, we use
interpretations of our GP Schema Theorem to obtain a GP
Building Block definition and to state a ``classical''
Building Block Hypothesis (BBH): that GP searches by
hierarchically combining building blocks. We report
that this approach is not convincing for several
reasons: it is difficult to find support for the
promotion and combination of building blocks solely by
rigourous interpretation of a GP Schema Theorem; even
if there were such support for a BBH, it is empirically
questionable whether building blocks always exist
because partial solutions of consistently above average
fitness and resilience to disruption are not assured;
also, a BBH constitutes a narrow and imprecise account
of GP search behaviour.
},
added-at = {2008-06-19T17:46:40.000+0200},
address = {Estes Park, Colorado, USA},
author = {O'Reilly, Una-May and Oppacher, Franz},
biburl = {https://www.bibsonomy.org/bibtex/2c6714f43338ea408cbbaedfb61f4fada/brazovayeye},
booktitle = {Foundations of Genetic Algorithms 3},
editor = {Whitley, L. Darrell and Vose, Michael D.},
interhash = {265dc6d0afee5fb717e1741d2add03cd},
intrahash = {c6714f43338ea408cbbaedfb61f4fada},
isbn = {1-55860-356-5},
keywords = {algorithms, genetic programming},
month = {31 July--2 August},
note = {Published 1995},
notes = {FOGA-3 An early version of this paper is an SFI
Technical Report (see \cite{OReilly:1992:tabbGP}). The
paper was accepted to a conference called ``Foundations
of Genetic Algorithms III''. It was presented at the
conference in Estes Park, Co, in July of 1993 (1994?)
and the latest version will appear in proceedings
forthcoming this year.
Presents a schema theorem for genetic programming based
on analogy with Holland's ST. Definition of schema more
general than Koza's (in Koza:book). Also delvelops
building block hyposthesis (again using analogy with
linear GA's). Then comprehensively trashes them both!
Some arguments against them apply equally well to
linear (bit string) GAs.
Chapter 4 of U.M. O'Reilly's PhD thesis
\cite{OReilly:thesis} is similar to this paper. Pages
192-196 of Chapter 7 summarize the results.},
organisation = {International Society for Genetic Algorithms},
pages = {73--88},
publisher = {Morgan Kaufmann},
publisher_address = {San Francisco, CA, USA},
size = {16 pages},
timestamp = {2008-06-19T17:49:03.000+0200},
title = {The Troubling Aspects of a Building Block Hypothesis
for Genetic Programming},
url = {http://citeseer.ist.psu.edu/oreilly92troubling.html},
year = 1994
}