Abstract
Historical markings in the NEAT algorithm provides a powerful feature for easy genetic alignment of any networks in the population, and allows speciation to protect networks with novelties. The original approach incorporated in NEAT always generates a new record for a connection with a unique ID when the connection is proposed in a generation. However, because of this mechanism, identical novelties developed in different generations are associated with different IDs and are not recognized as matching connections between networks. Despite popularity of the NEAT algorithm, there has been no existing study, which empirically investigates impact of this encoding on behavioral dynamics. The aim of this study is: firstly, to theoretically discuss generation context-dependent and generation context-free definitions for innovations (GC vs. GC-F); secondly, experimentally compare them on an XOR experiment under different speciation scenarios.
Our analyses suggest that the GC-F approach produces 40–50% less innovation records than the GC approach. Moreover, the GC algorithm exploits more innovation records from the register. However, the assumption about a higher number of species for the GC approach is observed to be true only for the first 30 generations. The difference represents a maximum of 10% decline of GC-F rates when compared to GC. The analysis of migratory patterns shows, that GC-F leads to higher migration to older species. However, differences in migration to younger species were minimal. In conclusion, the exuberant number of innovation records in the register in NEAT does not lead to critical behavioral differences.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Clune, J., Beckmann, B.E., Ofria, C., Pennock, R.T.: Evolving coordinated quadruped gaits with the hyperneat generative encoding. In: IEEE Congress on Evolutionary Computation, CEC 2009, pp. 2764–2771. IEEE (2009)
Clune, J., Mouret, J.B., Lipson, H.: The evolutionary origins of modularity. Proc. R. Soc. B. 280, 2012–2863 (2013). The Royal Society
Gauci, J., Stanley, K.O.: Autonomous evolution of topographic regularities in artificial neural networks. Neural Comput. 22(7), 1860–1898 (2010)
Hastings, E.J., Guha, R.K., Stanley, K.O.: Evolving content in the galactic arms race video game. In: IEEE Symposium on Computational Intelligence and Games, CIG 2009, pp. 241–248. IEEE (2009)
Kashtan, N., Alon, U.: Spontaneous evolution of modularity and network motifs. Proc. Natl. Acad. Sci. U.S.A. 102(39), 13773–13778 (2005)
Lehman, J., Stanley, K.O.: Evolving a diversity of virtual creatures through novelty search and local competition. In: Proceedings of the 13th Annual Conference on Genetic and Evolutionary Computation, pp. 211–218. ACM (2011)
Mayr, C.: Neat matlab. Last accessed (2017)
Mengistu, H., Huizinga, J., Mouret, J.B., Clune, J.: The evolutionary origins of hierarchy. PLoS Comput. Biol. 12(6), e1004829 (2016)
Stanley, K.O., Bryant, B.D., Miikkulainen, R.: Evolving adaptive neural networks with and without adaptive synapses. In: The 2003 Congress on Evolutionary Computation, CEC 2003, vol. 4, pp. 2557–2564. IEEE (2003)
Stanley, K.O., Bryant, B.D., Miikkulainen, R.: Real-time neuroevolution in the nero video game. IEEE Trans. Evol. Comput. 9(6), 653–668 (2005)
Stanley, K.O., D’Ambrosio, D.B., Gauci, J.: A hypercube-based encoding for evolving large-scale neural networks. Artif. Life 15(2), 185–212 (2009)
Stanley, K.O., Miikkulainen, R.: Evolving neural networks through augmenting topologies. Evol. Comput. 10(2), 99–127 (2002)
Stanley, K.O., Miikkulainen, R.: A taxonomy for artificial embryogeny. Artif. Life 9(2), 93–130 (2003)
Stanley, K.O., Miikkulainen, R.: Competitive coevolution through evolutionary complexification (2004)
Tan, M., Deklerck, R., Jansen, B., Bister, M., Cornelis, J.: A novel computer-aided lung nodule detection system for ct images. Med. Phys. 38(10), 5630–5645 (2011)
Trujillo, L., Muñoz, L., Naredo, E., Martínez, Y.: NEAT, there’s no bloat. In: Nicolau, M., Krawiec, K., Heywood, M.I., Castelli, M., García-Sánchez, P., Merelo, J.J., Rivas Santos, V.M., Sim, K. (eds.) EuroGP 2014. LNCS, vol. 8599, pp. 174–185. Springer, Heidelberg (2014). https://doi.org/10.1007/978-3-662-44303-3_15
Whiteson, S., Stone, P.: Evolutionary function approximation for reinforcement learning. J. Mach. Learn. Res. 7, 877–917 (2006)
Acknowledgments
This research is based upon works supported by Science Foundation Ireland under grant 13/RC/2094 which is co-funded under the European Regional Development Fund through the Southern & Eastern Regional Operational Programme to Lero - the Irish Software Research Centre (www.lero.ie).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2017 Springer International Publishing AG
About this paper
Cite this paper
Pastorek, L., O’Neill, M. (2017). Historical Markings in Neuroevolution of Augmenting Topologies Revisited. In: Martín-Vide, C., Neruda, R., Vega-Rodríguez, M. (eds) Theory and Practice of Natural Computing. TPNC 2017. Lecture Notes in Computer Science(), vol 10687. Springer, Cham. https://doi.org/10.1007/978-3-319-71069-3_19
Download citation
DOI: https://doi.org/10.1007/978-3-319-71069-3_19
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-71068-6
Online ISBN: 978-3-319-71069-3
eBook Packages: Computer ScienceComputer Science (R0)