ABSTRACT
We study if and when the inequality dp(H) ≤ relΔ(H) holds for schemas H in chromosomes that are structured as trees. The disruption probability dp(H) is the probability that a random cut of a tree limb will separate two fixed nodes of H. The relative diameter relΔ(H) is the ratio (max distance between two fixed nodes in H) / (max distance between two tree nodes), and measures how close together are the fixed nodes of H. Inequality dp(H) ≤ relΔ(H) is of significance in proving Schema Theorems for non-linear chromosomes, and so bears upon the success we can expect from genetic algorithms. For linear chromosomes, dp(H) ≤relΔ(H). Our results include the following. There is no constant c such that dp(H) ≤ c • relΔ(H) holds for arbitrary schemas and trees. This is illustrated in trees with eccentric, stringy shapes. Matters improve for dense, ball-like trees, explained herein. Inequality dp(H) ≤ relΔ(H) always holds in such trees, except for certain atypically large schemas. Thus, the more compact are our tree-structured chromo-somes, the better we can expect our genetic algorithms to work.
- Goldberg, David E. (1989). Genetic Algorithms in Search, Optimization, and Machine Learning. Addison-Wesley Publishing, Reading, MA. Google ScholarDigital Library
- Greene, William A. (2000). "A Non-Linear Schema Theorem for Genetic Algorithms," in Whitley, D. et al. (Eds.) Proceedings of the Genetic and Evolutionary Computation Conference (GECCO-2000), pages 189--194. Morgan Kaufmann Publishers, San Francisco, CA.Google Scholar
- Greene, William A. (2004) "Schema Disruption in Chromosomes that are Structured as Binary Trees", in K. Deb et al. (eds.), Proceedings of the Genetic and Evolutionary Computation Congress, June 26-30, 2004, Seattle, WA (pages 1197--1207). Springer Verlag LNCS 3102, New York, NY.Google Scholar
- Holland, John (1975). Adaptation in Natural and Artificial Systems. University of Michigan Press, Ann Arbor, MI. Google ScholarDigital Library
- Koza, John R. (1992). Genetic Programming: On the Programming of Computers by Natural Selection. MIT Press, Cambridge, MA Google ScholarDigital Library
- O'Reilly, Una-May (1995). An Analysis of Genetic Programming. PhD thesis, Carleton University, Ottawa-Carleton Institute for Computer Science, Ottawa, Ontario, Canada, 22 September 1995. Google ScholarDigital Library
- O'Reilly, Una-May, and Franz Oppacher (1995). "The Troubling Aspects of a Building Block Hypothesis for Genetic Programming", in Whitley, D. and Vose, M. D. (eds.) Foundations of Genetic Algorithms 3. Morgan Kaufmann Publishers, San Francisco.Google Scholar
- Poli, Riccardo, and William Langdon (1998). "Schema Theory for Genetic Programming with One-Point Crossover and Point Mutation". Evolutionary Computation 6(3), pages 231--252. MIT Press, Cambridge, MA. Google ScholarDigital Library
- Rosca, Justinian P. (1997). "Analysis of Complexity Drift in Genetic Programming", in Koza, John R. et al. (eds) Genetic Programming 1997: Proceedings of the Second Annual Conference (pages 286--294). Morgan Kaufmann Publishers, San Francisco, CA.Google Scholar
- Whigham, Peter A. (1995) "A Schema Theorem for Context-Free Grammars", in 1995 IEEE Conference on Evolutionary Computation, Vol 1, pages 178--181. IEEE Press.Google ScholarCross Ref
Index Terms
- Schema disruption in tree-structured chromosomes
Recommendations
General schema theory for genetic programming with subtree-swapping crossover: Part II
This paper is the second part of a two-part paper which introduces a general schema theory for genetic programming (GP) with subtree-swapping crossover (Part I (Poli and McPhee, 2003)). Like other recent GP schema theory results, the theory gives an ...
Constructing a minimum height elimination tree of a tree in linear time
Given a graph, finding an optimal vertex ranking and constructing a minimum height elimination tree are two related problems. However, an optimal vertex ranking does not by itself provide enough information to construct an elimination tree of minimum ...
Exact Schema Theory and Markov Chain Models for Genetic Programming and Variable-length Genetic Algorithms with Homologous Crossover
Genetic Programming (GP) homologous crossovers are a group of operators, including GP one-point crossover and GP uniform crossover, where the offspring are created preserving the position of the genetic material taken from the parents. In this paper we ...
Comments