From black-box complexity to designing new genetic algorithms

Black-box complexity theory recently produced several surprisingly fast black-box optimization algorithms. In this work, we exhibit one possible reason: These black-box algorithms often profit from solutions inferior to the previous-best. In contrast, evolutionary approaches guided by the “survival...

Full description

Saved in:
Bibliographic Details
Published in:Theoretical computer science Vol. 567; pp. 87 - 104
Main Authors: Doerr, Benjamin, Doerr, Carola, Ebel, Franziska
Format: Journal Article
Language:English
Published: Elsevier B.V 16.02.2015
Elsevier
Subjects:
Algorithms
Asymptotic properties
Black-box complexity
Computer Science
Crossovers
Evolutionary
Genetic algorithms
Heuristic search
Mathematical analysis
Mathematical models
Mutations
Neural and Evolutionary Computing
Run time (computers)
Runtime analysis
Theory of randomized search heuristics
Heuristic search
Theory of randomized search heuristics
Runtime analysis
Genetic algorithms
Black-box complexity
ISSN:0304-3975, 1879-2294
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Black-box complexity theory recently produced several surprisingly fast black-box optimization algorithms. In this work, we exhibit one possible reason: These black-box algorithms often profit from solutions inferior to the previous-best. In contrast, evolutionary approaches guided by the “survival of the fittest” paradigm often ignore such solutions. We use this insight to design a new crossover-based genetic algorithm. It uses mutation with a higher-than-usual mutation probability to increase the exploration speed and crossover with the parent to repair losses incurred by the more aggressive mutation. A rigorous runtime analysis proves that our algorithm for many parameter settings is asymptotically faster on the OneMax test function class than all what is known for classic evolutionary algorithms. A fitness-dependent choice of the offspring population size provably reduces the expected runtime further to linear in the dimension. Our experimental analysis on several test function classes shows advantages already for small problem sizes and broad parameter ranges. Also, a simple self-adaptive choice of these parameters gives surprisingly good results.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISSN:0304-3975
1879-2294
DOI:10.1016/j.tcs.2014.11.028