Incremental co-evolution of organisms: A new approach for optimization and discovery of strategies
In the field of optimization and machine learning techniques, some very efficient and promising tools like Genetic Algorithms (GAs) and Hill-Climbing have been designed. In this same field, the Evolving Non-Determinism (END) model presented in this paper proposes an inventive way to explore the space of states that, using the simulated “incremental” co-evolution of some organisms, remedies some drawbacks of these previous techniques and even allow this model to outperform them on some difficult problems.
This new model has been applied to the sorting network problem, a reference problem that challenged many computer scientists, and an original one-player game named Solitaire. For the first problem, the END model has been able to build from “scratch” some sorting networks as good as the best known for the 16-input problem. It even improved by one comparator a 25 years old result for the 13-input problem. For the Solitaire game, END evolved a strategy comparable to a human designed strategy.
KeywordsEvolutionary optimization Simulated co-evolution Sorting networks
Unable to display preview. Download preview PDF.
- Richard Belew and Thomas Kammeyer: Evolving Aesthetic Sorting Networks using Developmental Grammars. In Proceedings of the Fifth International Conference of Genetic Algorithms.Google Scholar
- Roberto Bisiani: Beam Search. In Encyclopedia of Artificial Intelligence, Vol. 2, Second Edition, John Wiley & Sons, 1992.Google Scholar
- Gary L. Drescher: Evolution of 16-Number Sorting Networks Revisited. Submitted.Google Scholar
- Milton W. Green: Some Improvements in Nonadaptive Sorting Algorithms. Stanford Research Institute. Menlo Park, California.Google Scholar
- W. Daniel Hillis: Co-Evolving Parasites Improve Simulated Evolution as an Optimization Procedure. In Artificial Life II, Langton, et al, Eds. Addison Wesley, 1992, pp. 313–324.Google Scholar
- Hugues Juillé: Evolving Non-Determinism: An Inventive and Efficient Tool for Optimization and Discovery of Strategies. Draft paper, Computer Science Departement, Brandeis University, 27 pp.Google Scholar
- Kim Kinnear: Generality and Difficulty in GP: Evolving a Sort. In Proceedings of the Fifth International Conference on Genetic Algorithms, S. Forrest; Morgan Kaufmann Publishers, 1993.Google Scholar
- Donald E. Knuth: The Art of Computer Programming: Volume 3 — Sorting and Searching. Addison Wesley, 1973.Google Scholar
- Steven Levy: Artificial Life: the Quest for a New Creation.Pantheon Books, 1992.Google Scholar
- Ian Parberryr: A Computer-Assisted Optimal Depth Lower Bound for Nine-Input Sorting Networks. In Mathematical Systems Theory, No 24, 1991, pp. 101–116.Google Scholar
- Conor Ryan: Pygmies and Civil Servants. In Advances in Genetic Programming, Kim Kinnear, Ed. MIT Press, 1994.Google Scholar
- Nicol N. Schraudolph and Richard K. Belew: Dynamic Parameter Encoding for Genetic Algorithms. In Machine Learning, Vol. 9, 1992, pp. 9–21.Google Scholar
- Patrick Tufts and Hugues Juillé: Evolving Non-Deterministic Algorithms for Efficient Sorting Networks. Submitted.Google Scholar