- author = "Stephen Kelly and Malcolm I. Heywood",
- title = "Knowledge Transfer from Keepaway Soccer to Half-field Offense through Program Symbiosis: Building Simple Programs for a Complex Task",
- booktitle = "GECCO '15: Proceedings of the 2015 Annual Conference on Genetic and Evolutionary Computation",
- year = "2015",
- editor = "Sara Silva and Anna I Esparcia-Alcazar and Manuel Lopez-Ibanez and Sanaz Mostaghim and Jon Timmis and Christine Zarges and Luis Correia and Terence Soule and Mario Giacobini and Ryan Urbanowicz and Youhei Akimoto and Tobias Glasmachers and Francisco {Fernandez de Vega} and Amy Hoover and Pedro Larranaga and Marta Soto and Carlos Cotta and Francisco B. Pereira and Julia Handl and Jan Koutnik and Antonio Gaspar-Cunha and Heike Trautmann and Jean-Baptiste Mouret and Sebastian Risi and Ernesto Costa and Oliver Schuetze and Krzysztof Krawiec and Alberto Moraglio and Julian F. Miller and Pawel Widera and Stefano Cagnoni and JJ Merelo and Emma Hart and Leonardo Trujillo and Marouane Kessentini and Gabriela Ochoa and Francisco Chicano and Carola Doerr",
- pages = "1143--1150",
- publisher = "ACM",
- publisher_address = "New York, NY, USA",
- month = "11-15 " # jul,
- organisation = "SIGEVO",
- address = "Madrid, Spain",
- keywords = "genetic algorithms, genetic programming, Integrative Genetic and Evolutionary Computation",
- isbn13 = "978-1-4503-3472-3",
-
URL = "
http://stephenkelly.ca/research_files/open-kelly15.pdf",
-
URL = "
http://doi.acm.org/10.1145/2739480.2754798",
-
DOI = "
10.1145/2739480.2754798",
- size = "8 pages",
- abstract = "Half-field Offense (HFO) is a sub-task of Robocup 2D Simulated Soccer. HFO is a challenging, multi-agent machine learning problem in which a team of offense players attempt to manoeuvre the ball past a defending team and around the goalie in order to score. The agent's sensors and actuators are noisy, making the problem highly stochastic and partially observable. These same real-world characteristics have made Keepaway soccer, which represents one sub-task of HFO, a popular testbed in the reinforcement learning and task-transfer literature in particular. We demonstrate how policies initially evolved for Keepaway can be reused within a symbiotic framework for coevolving policies in genetic programming (GP), with no additional training or transfer function, in order to improve learning in the HFO task. Moreover, the highly modular policies discovered by GP are shown to be significantly less complex than solutions based on traditional value-function optimization while achieving the same level of play in HFO.",
-
notes = "open-kelly15.pdf reformated
Also known as \cite{2754798} GECCO-2015 A joint meeting of the twenty fourth international conference on genetic algorithms (ICGA-2015) and the twentith annual genetic programming conference (GP-2015)",
