Does Relaxing Strict Acceptance Condition Improve Test Based Pareto Coevolution?

Abstract

The strict acceptance condition between parent and child is one of the guarantees of monotonic progress in Pareto co-evolution. However, this condition results in "stalling" in progress especially when the cardinality of test set is as large as in the scale of population size in a test based coevolutionary algorithm. We presented two variants of Pareto based Coevolutionary Hill Climber algorithms - rP-PHC-P that relax the strict condition based on competitive shared fitness of non-comparable (parent, child) pairs and fP-PHC-P which relaxes the condition in the form of discarding Pareto dominated candidate solutions regardless of being parent or child as soon as the Pareto front is created. Both of the algorithms improve the progress, are successfully tested for avoiding overspecialization and keep less non-unique candidate solutions in the population slot compare to P-PHC-P.

Publication Date

2-2-2018

Publication Title

2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings

Volume

2018-January

Number of Pages

1-8

Document Type

Article; Proceedings Paper

Personal Identifier

scopus

DOI Link

https://doi.org/10.1109/SSCI.2017.8285424

Socpus ID

85046130419 (Scopus)

Source API URL

https://api.elsevier.com/content/abstract/scopus_id/85046130419

This document is currently not available here.

Share

COinS