EA Models of Population Fixed Points Versus Mutations Rates for Functions of Unitation

posted 11/23/05

J. Neal Richter  
Computer Science Department,
Montana State University
Bozeman, Montana
richter@cs.montana.edu

John Paxton  
Computer Science Department,
Montana State University
Bozeman, Montana
paxton@cs.montana.edu

Alden H. Wright
Computer Science Dept.  
Univ. of Montana Missoula, MT 59812 USA
wright@cs.umt.edu
(406) 243-4790


Abstract

Using a dynamic systems model for the Simple Genetic Algorithm 
due to Vose, we analyze the fixed point behavior of the model 
without crossover applied to functions of unitation.  Unitation 
functions are simplified fitness functions that reduce the search 
space into a smaller number of equivalence classes.  This reduction 
allows easier computation of fixed points.  We also create a dynamic 
systems model from a simple nondecreasing EA like the (1+1) EA and 
variants, then analyze this models on unitation classes.