PT Unknown
AU David Roche
   Debora Gil
   Jesus Giraldo
TI Detecting loss of diversity for an efficient termination of EAs
BT 15th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing
PY 2013
BP 561
EP 566
DI 10.1109/SYNASC.2013.79
DE EA termination; EA population diversity; EA steady state
AB Termination of Evolutionary Algorithms (EA) at its steady state so that useless iterations are not performed is a main point for its efficient application to black-box problems. Many EA algorithms evolve while there is still diversity in their population and, thus, they could be terminated by analyzing the behavior some measures of EA population diversity. This paper presents a numeric approximation to steady states that can be used to detect the moment EA population has lost its diversity for EA termination. Our condition has been applied to 3 EA paradigms based on diversity and a selection of functionscovering the properties most relevant for EA convergence.Experiments show that our condition works regardless of the search space dimension and function landscape.
ER