Nullifying the Inherent Bias of Non-invariant Exploratory Landscape Analysis Features

Raphael Patrick Prager*, Heike Trautmann

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

14 Citations (Scopus)
102 Downloads (Pure)

Abstract

Exploratory landscape analysis (ELA) in single-objective black-box optimization relies on a comprehensive and large set of numerical features characterizing problem instances. Those foster problem understanding and serve as basis for constructing automated algorithm selection models choosing the best suited algorithm for a problem at hand based on the aforementioned features computed prior to optimization. This work specifically points to the sensitivity of a substantial proportion of these features to absolute objective values, i.e., we observe a lack of shift and scale invariance. We show that this unfortunately induces bias within automated algorithm selection models, an overfitting to specific benchmark problem sets used for training and thereby hinders generalization capabilities to unseen problems. We tackle these issues by presenting an appropriate objective normalization to be used prior to ELA feature computation and empirically illustrate the respective effectiveness focusing on the BBOB benchmark set.

Original languageEnglish
Title of host publicationApplications of Evolutionary Computation
Subtitle of host publication26th European Conference, EvoApplications 2023, Held as Part of EvoStar 2023, Brno, Czech Republic, April 12–14, 2023, Proceedings
EditorsJoão Correia, Stephen Smith, Raneem Qaddoura
Place of PublicationCham
PublisherSpringer
Pages411-425
Number of pages15
ISBN (Electronic)978-3-031-30229-9
ISBN (Print)978-3-031-30228-2
DOIs
Publication statusPublished - 2023
Event26th International Conference on Applications of Evolutionary Computation, EvoApplications 2023 - Brno, Czech Republic
Duration: 12 Apr 202314 Apr 2023
Conference number: 26

Publication series

NameLecture Notes in Computer Science
PublisherSpringer
Volume13989
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference26th International Conference on Applications of Evolutionary Computation, EvoApplications 2023
Abbreviated titleEvoApplications
Country/TerritoryCzech Republic
CityBrno
Period12/04/2314/04/23
Otherheld as part of EvoStar 2023

Keywords

  • Automated algorithm selection
  • Exploratory landscape analysis
  • Invariance
  • 2024 OA procedure

Fingerprint

Dive into the research topics of 'Nullifying the Inherent Bias of Non-invariant Exploratory Landscape Analysis Features'. Together they form a unique fingerprint.

Cite this