Enzyme Genetic Programming - Cellular Computing - page 40

CONCLUSIONS
This chapter presented an overview of enzyme genetic programming, an ap-
proach to evolutionary computation motivated by the metabolic processing of
cells. Enzyme GP can be distinguished from conventional GP by its use of a
program representation and developmental process derived from biology. The
aim of the approach is to capture the elements of biological representations
that contribute to their evolvability and adapt these for artificial evolution. The
resulting system represents an executable structure as a collection of enzymelike
computational elements that interact with one another according to their own
interaction preferences.
The method has been applied to a number of problems in the domain of
digital circuit design but does not yet indicate any significant performance ad-
vantage over other GP approaches. However, analysis of solution-size evolution
shows that, unlike most other GP approaches, enzyme GP does not suffer from
bloat. On the contrary, enzyme GP appears to be biased toward finding smaller
solutions to problems, and it achieves this without any form of fitness penalty
or operator modification.
Enzyme GP has a number of interesting properties. Perhaps most important
of these is that the context of each component within a program is independent
of its position within the genome. Furthermore, the context of a component is
recorded using a description independent of any particular program and, con-
sequently, the role of a component can be preserved following recombination.
References
[1] L. Altenberg. Emergent phenomena in genetic programming. In A. V. Sebald and
L. J. Fogel, editors, Evolutionary Programming—Proceedings of the Third Annual
Conference , pages 233-241. World Scientific Publishing, River Edge, NJ, 1994.
[2] Lee Altenberg. The evolution of evolvability in genetic programming. In K. Kin-
near, Jr., editor, Advances in Genetic Programming , pages 47-74. MIT Press,
Cambridge, 1994.
[3] P. Angeline. Subtree crossover: building block engine or macromutation? In
J. Koza, editor, Genetic Programming 1997: Proceedings of the Second Annual
Conference, GP97 , pages 240-248. Morgan Kaufmann, San Francisco, CA, 1997.
[4] P. Angeline. Multiple interacting programs: A representation for evolving complex
behaviors. Cybernet. Syst. , 29(8):779-806, 1998.
[5] T. Back. Evolutionary Algorithms in Theory and Practice . Oxford University Press,
Oxford, 1996.
[6] Wolfgang Banzhaf, Peter Nordin, Robert E.Keller, and Frank D. Francone. Genetic
Programming—An Introduction . Morgan Kaufmann, San Francisco, CA, 1998.
[7] T. Blickle and L. Thiele. Genetic programming and redundancy. In J. Hopf, editor,
Genetic Algorithms within the Framework of Evolutionary Computation (Work-