Multi-modal valley-adaptive memetic algorithm for efficient discovery of first-order saddle points

Mostafa Ellabaan*, Xianshun Chen, Nguyen Quang Huy

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

Abstract

First-order saddle point represents an important landmark on the problem landscape. This point lies along the minimum energy path connecting two minima, more specifically at the point with maximum energy on the path. Unlike minima or maxima, to identify first-order saddle points require both maximization and minimization tasks. Finding such points is extremely difficult. In this paper, we present a real-coded memetic algorithm for locating first-order saddle points. The proposed algorithm leverage the advantage of valley- adaptive clearing scheme in maintaining multiple solutions and Schlegel algorithm in achieving fast and precise convergence. Empirical results shown that the proposed algorithms achieve more than 90% with converge speed of more than 100 fold when comparing to its evolutionary compeers.

Original languageEnglish
Title of host publicationSimulated Evolution and Learning - 9th International Conference, SEAL 2012, Proceedings
Volume7673 LNCS
Publication date2012
Pages83-92
ISBN (Print)9783642348587
DOIs
Publication statusPublished - 2012
Externally publishedYes
Event9th International Conference on Simulated Evolution and Learning, SEAL 2012 - Hanoi, Viet Nam
Duration: 16 Dec 201219 Dec 2012

Conference

Conference9th International Conference on Simulated Evolution and Learning, SEAL 2012
CountryViet Nam
CityHanoi
Period16/12/201219/12/2012
SeriesLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume7673 LNCS
ISSN0302-9743

Keywords

  • Memetic algorithm
  • Multi-Modal optimization
  • Saddle points

Cite this

Ellabaan, M., Chen, X., & Huy, N. Q. (2012). Multi-modal valley-adaptive memetic algorithm for efficient discovery of first-order saddle points. In Simulated Evolution and Learning - 9th International Conference, SEAL 2012, Proceedings (Vol. 7673 LNCS, pp. 83-92). Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol.. 7673 LNCS https://doi.org/10.1007/978-3-642-34859-4_9