TI Genetic Invariance:  A New Approach to Genetic Algorithms 

AV ftp thorhild.cs.ualberta.ca:pubTR91-03.ps.Z 

AV mail britta@cs.ualberta.ca  

OR ALBRT 

LT TR 91-03 

AU Michael Lewchuk 

AU Joseph C. Culberson 

MN February 

YR 1991 

AB ABS We present a different selection system for genetic
algorithms using overlapping populations which has the property
that the total genetic content of the population does not change.
Instead, at least for the simple function studied, optimization
is achieved by a separation of "good" characteristics from "poor"
ones. Thus this method does not require the use of mutation to
prevent premature convergence, since convergence in the usual
sense never occurs. We present some analyses on a simple function
as a preliminary study of this technique.