- author = "E. Howard N. Oakley",
- title = "Genetic Programming, the Reflection of Chaos, and the Bootstrap: Towards a useful Test for Chaos",
- booktitle = "Genetic Programming 1996: Proceedings of the First Annual Conference",
- editor = "John R. Koza and David E. Goldberg and David B. Fogel and Rick L. Riolo",
- year = "1996",
- month = "28--31 " # jul,
- keywords = "genetic algorithms, genetic programming",
- pages = "175--181",
- address = "Stanford University, CA, USA",
- publisher = "MIT Press",
-
URL = "
http://www.quercus.demon.co.uk/oakgp96.pdf",
-
URL = "http://cognet.mit.edu/sites/default/files/books/9780262315876/pdfs/9780262315876_chap21.pdf",
-
URL = "
http://cognet.mit.edu/library/books/view?isbn=0262611279",
-
DOI = "
doi:10.7551/mitpress/3242.003.0024",
- size = "7 pages",
-
abstract = "This study assessed the use of genetic programming
(GP) to diagnose chaos. Fifty GP runs were performed on
chaotic data, generated from the Mackey-Glass delay
differential equation, on one surrogate with the same
Fourier power spectrum and statistics but without
chaotic dynamics, and on a random walk series. Single
runs were performed on 50 different surrogates of the
chaotic series. Fitness was measured across 5 separate
forecast periods of 65 points each, each based upon 10
prior input data points.
Fittest program fragments for the chaotic series evolved later and were more complicated than those for the surrogates. Relative to fitnesses achieved by constant linear predictions, fitnesses from the chaotic series were also better. Random walk data resulted in an impoverished GP process, with quick evolution of simple program fragments but no later evolutionary improvement. This comparative test merits assessment on other datasets, and its implications with respect to the statistical bootstrap and GP estimation are discussed.",
- notes = "GP-96",
