Smart Media Journal (스마트미디어저널)

THE KOREAN INSTITUTE OF SMART MEDIA (한국스마트미디어학회)
권호 표지
DOI QR코드

DOI QR Code

An Improved MAP-Elites Algorithm via Rotational Invariant Operator in Differential Evolution for Continuous Optimization

연속 최적화를 위한 개선된 MAP-Elites 알고리즘

  • 최태종 (전남대학교 데이터사이언스대학원)
  • Received : 2023.12.04
  • Accepted : 2024.01.13
  • Published : 2024.02.29
Download PDF
⟨ Previous Next ⟩

Abstract

In this paper, we propose a new approach that enhances the continuous optimization performance of the MAP-Elites algorithm. The existing self-referencing MAP-Elites algorithm employed the "DE/rand/1/bin" operator from the differential evolution algorithm, which, due to its lack of rotational invariance, led to a degradation in optimization performance when there were high correlations among variables. The proposed algorithm replaces the "DE/rand/1/bin" operator with the "DE/current-to-rand/1" operator. This operator, possessing rotational invariance, ensures robust performance even in cases where there are high correlations among variables. Experimental results confirm that the proposed algorithm performs better than the comparison algorithms.

이 연구에서는 MAP-Elites 알고리즘의 연속 최적화 성능을 향상한 새로운 접근법을 제안한다. 기존의 자기 참조 MAP-Elites 알고리즘은 차분 진화 알고리즘의 "DE/rand/1/bin" 연산자를 사용했는데, 이 연산자는 회전 불변이 아니라서 각 변수 간의 상관관계가 높은 경우 성능이 감소하는 문제가 존재한다. 제안하는 알고리즘은 "DE/rand/1/bin" 연산자 대신에 "DE/current-to-rand/1" 연산자를 사용한다. 이 연산자는 회전 불변성을 가지므로 각 변수 간의 상관관계가 높은 분리 불가능 최적화 문제에서도 강건한 성능을 보장할 수 있다. 실험 결과, 제안하는 알고리즘이 비교 알고리즘들에 비해 높은 성능을 발휘함을 확인했다.

Keywords

Acknowledgement

이 성과는 정부(과학기술정보통신부)의 재원으로 한국연구재단의 지원을 받아 수행된 연구임(RS-2023-00214326). 이 성과는 정부(과학기술정보통신부)의 재원으로 한국연구재단의 지원을 받아 수행된 연구임(RS-2023-00242528).

References

  1. Pugh, Justin K., Lisa B. Soros, and Kenneth O. Stanley. "Quality diversity: A new frontier for evolutionary computation.," Frontiers in Robotics and AI, vol. 3, pp. 40(1-17), 2016. https://doi.org/10.3389/frobt.2016.00040
  2. Mouret, Jean-Baptiste, and Jeff Clune. "Illuminating search spaces by mapping elites," arXiv preprint arXiv:1504.04909, 2015.
  3. Vassiliades, Vassilis, Konstantinos Chatzilygeroudis, and Jean-Baptiste Mouret. "Using centroidal voronoi tessellations to scale up the multidimensional archive of phenotypic elites algorithm," IEEE Transactions on Evolutionary Computation, vol. 22, no. 4, pp. 623-630, 2017. https://doi.org/10.1109/TEVC.2017.2735550
  4. Choi, Tae Jong, and Julian Togelius. "Self-referential quality diversity through differential MAP-Elites," Proceedings of the Genetic and Evolutionary Computation Conference, pp. 502-509, 2021.
  5. Storn, Rainer, and Kenneth Price. "Differential evolution-a simple and efficient heuristic for global optimization over continuous spaces," Journal of global optimization, vol. 11, pp. 341-359, 1997. https://doi.org/10.1023/A:1008202821328
  6. Lehman, Joel, and Kenneth O. Stanley. "Abandoning objectives: Evolution through the search for novelty alone," Evolutionary computation, vol. 19, no. 2, pp. 189-223, 2011. https://doi.org/10.1162/EVCO_a_00025
  7. Lehman, Joel, and Kenneth O. Stanley. "Evolving a diversity of virtual creatures through novelty search and local competition," Proceedings of the 13th annual conference on Genetic and evolutionary computation, pp. 211-218, 2011.
  8. Fontaine, Matthew C., Scott Lee, Lisa B. Soros, Fernando de Mesentier Silva, Julian Togelius, and Amy K. Hoover. "Mapping hearthstone deck spaces through map-elites with sliding boundaries," Proceedings of The Genetic and Evolutionary Computation Conference, pp. 161-169, 2019.
  9. Cully, Antoine, and Yiannis Demiris. "Quality and diversity optimization: A unifying modular framework," IEEE Transactions on Evolutionary Computation, vol. 22, no. 2, pp. 245-259, 2017. https://doi.org/10.1109/TEVC.2017.2704781
  10. Fontaine, Matthew C., Julian Togelius, Stefanos Nikolaidis, and Amy K. Hoover. "Covariance matrix adaptation for the rapid illumination of behavior space.," Proceedings of the 2020 genetic and evolutionary computation conference, pp. 94-102, 2020.
  11. Kent, Paul, and Juergen Branke. "Bop-elites, a bayesian optimisation algorithm for quality-diversity search," arXiv preprint arXiv:2005.04320, 2020.
  12. N. H. Awad, M. Z. Ali, J. J. Liang, B. Y. Qu and P. N. Suganthan, "Problem Definitions and Evaluation Criteria for the CEC 2017 Special Session and Competition on Single Objective Bound Constrained Real-Parameter Numerical Optimization," Technical Report, Nanyang Technological University, Singapore, 2016.