- author = "Joerg Stork and Martin Zaefferer and Thomas Bartz-Beielstein and A. E. Eiben",
- title = "Surrogate Models for Enhancing the Efficiency of Neuroevolution in Reinforcement Learning",
- booktitle = "GECCO '19: Proceedings of the Genetic and Evolutionary Computation Conference",
- year = "2019",
- editor = "Manuel Lopez-Ibanez and Thomas Stuetzle and Anne Auger and Petr Posik and Leslie {Peprez Caceres} and Andrew M. Sutton and Nadarajen Veerapen and Christine Solnon and Andries Engelbrecht and Stephane Doncieux and Sebastian Risi and Penousal Machado and Vanessa Volz and Christian Blum and Francisco Chicano and Bing Xue and Jean-Baptiste Mouret and Arnaud Liefooghe and Jonathan Fieldsend and Jose Antonio Lozano and Dirk Arnold and Gabriela Ochoa and Tian-Li Yu and Holger Hoos and Yaochu Jin and Ting Hu and Miguel Nicolau and Robin Purshouse and Thomas Baeck and Justyna Petke and Giuliano Antoniol and Johannes Lengler and Per Kristian Lehre",
- isbn13 = "978-1-4503-6111-8",
- pages = "934--942",
- address = "Prague, Czech Republic",
-
DOI = "
doi:10.1145/3321707.3321829",
- publisher = "ACM",
- publisher_address = "New York, NY, USA",
- month = "13-17 " # jul,
- organisation = "SIGEVO",
- keywords = "genetic algorithms, genetic programming, Cartesian Genetic Programming, CGP, CGP-ANN, ANN, Evolutionary algorithms, Gaussian processes, Computing methodologies, Neural networks, Neuroevolution, Surrogate Models, Reinforcement Learning, Kriging",
- size = "9 pages",
- abstract = "In the last years, reinforcement learning received a lot of attention. One method to solve reinforcement learning tasks is Neuroevolution, where neural networks are optimized by evolutionary algorithms. A disadvantage of Neuroevolution is that it can require numerous function evaluations, while not fully using the available information from each fitness evaluation. This is especially problematic when fitness evaluations become expensive. To reduce the cost of fitness evaluations, surrogate models can be employed to partially replace the fitness function. The difficulty of surrogate modelling for Neuroevolution is the complex search space and how to compare different networks. To that end, recent studies showed that a kernel based approach, particular with phenotypic distance measures, works well. These kernels compare different networks via their behaviour (phenotype) rather than their topology or encoding (genotype). In this work, we discuss the use of surrogate model-based Neuroevolution (SMB-NE) using a phenotypic distance for reinforcement learning. In detail, we investigate a) the potential of SMB-NE with respect to evaluation efficiency and b) how to select adequate input sets for the phenotypic distance measure in a reinforcement learning problem. The results indicate that we are able to considerably increase the evaluation efficiency using dynamic input sets.",
-
notes = "SMB-NE, CartPole-v1 and MountainCar-v0, DynSet
10.
Also known as \cite{3321829} GECCO-2019 A Recombination of the 28th International Conference on Genetic Algorithms (ICGA) and the 24th Annual Genetic Programming Conference (GP)",
