When Will a Genetic Algorithm Outperform Hill-Climbing?

Melanie Mitchell and John H. Holland

Working Papers from Santa Fe Institute

Abstract: In this paper we review some previously published experimental results in which a simple hill-climbing algorithm---Random Mutation Hill-Climbing (RMHC)---significantly outperforms a genetic algorithm on a simple ``Royal Road'' function. We present an analysis of RMHC followed by an analysis of an ``idealized'' genetic algorithm (IGA) that is in turn significantly faster than RMHC. We isolate the features of the IGA that allow for this speedup, and discuss how these features can be incorparated into a real GA and a fitness landscape, making the GA better approximate the IGA. We use these features to design a modified version of the previously published experiments, and give new experimental results comparing the GA and RMHC.

Date: 1993-06
References: Add references at CitEc
Citations: View citations in EconPapers (2)

There are no downloads for this item, see the EconPapers FAQ for hints about obtaining it.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:wop:safiwp:93-06-037

Access Statistics for this paper

More papers in Working Papers from Santa Fe Institute Contact information at EDIRC.
Bibliographic data for series maintained by Thomas Krichel ().