Effective fitness

In natural evolution and artificial evolution (e.g. artificial life and evolutionary computation) the fitness (or performance or objective measure) of a schema is rescaled to give its effective fitness which takes into account crossover and mutation. That is effective fitness can be thought of as the fitness that the schema would need to have in order to increase or decrease as a fraction of the population as it actually does with crossover and mutation present but as if they were not.

Part of a series on the
Evolutionary algorithm
Genetic algorithm
Genetic programming

Problem solving with evolutionary computation is realized with a cost function.[1] If cost functions are applied to swarm optimization they are called a fitness function. Strategies like reinforcement learning[2] and NEAT neuroevolution[3] are creating a fitness landscape which describes the reproductive success of cellular automata.[4][5]

A normal fitness function fits to a problem,[6] while an effective fitness function is an assumption if the objective was reached.[7] The difference is important for designing fitness functions with algorithms like novelty search in which the objective of the agents is unknown.[8][9]

References
  1. J. David Schaffer and Heike Sichtig and Craig Laramee (2009). A series of failed and partially successful fitness functions for evolving spiking neural networks. Proceedings of the 11th annual conference companion on Genetic and evolutionary computation conference - GECCO 09. ACM Press. doi:10.1145/1570256.1570378.
  2. Afanasyeva, Arina and Buzdalov, Maxim (2012). Optimization with auxiliary criteria using evolutionary algorithms and reinforcement learning. Proceedings of 18th International Conference on Soft Computing MENDEL 2012. 2012. pp. 58–63.CS1 maint: multiple names: authors list (link)
  3. Mohammad Divband Soorati and Heiko Hamann (2015). The Effect of Fitness Function Design on Performance in Evolutionary Robotics. Proceedings of the 2015 on Genetic and Evolutionary Computation Conference - GECCO 15. ACM Press. doi:10.1145/2739480.2754676.
  4. Peter F. Stadler and Christopher R. Stephens (2003). "Landscapes and Effective Fitness". Comments on Theoretical Biology. Informa UK Limited. 8 (4–5): 389–431. doi:10.1080/08948550302439.
  5. Bagnoli, Franco (1998). "Cellular automata". arXiv:cond-mat/9810012.
  6. Fernandez, Aaron Carl T (2017). "Creating a fitness function that is the right fit for the problem at hand". Cite journal requires |journal= (help)
  7. Hisashi Handa (2006). Fitness function for finding out robust solutions on time-varying functions. Proceedings of the 8th annual conference on Genetic and evolutionary computation GECCO 06. ACM Press. doi:10.1145/1143997.1144186.
  8. Joel Lehman and Kenneth O. Stanley (2011). "Abandoning Objectives: Evolution Through the Search for Novelty Alone". Evolutionary Computation. MIT Press - Journals. 19 (2): 189–223. doi:10.1162/evco_a_00025. PMID 20868264.
  9. Woolley, Brian G and Stanley, Kenneth O (2012). "Exploring promising stepping stones by combining novelty search with interactive evolution". arXiv:1207.6682 [cs.NE].CS1 maint: multiple names: authors list (link)


This article is issued from Wikipedia. The text is licensed under Creative Commons - Attribution - Sharealike. Additional terms may apply for the media files.