- author = "Jeannie Fitzgerald and R. Muhammad Atif Azad and Conor Ryan",
- title = "A bootstrapping approach to reduce over-fitting in genetic programming",
- booktitle = "GECCO '13 Companion: Proceeding of the fifteenth annual conference companion on Genetic and evolutionary computation conference companion",
- year = "2013",
- editor = "Christian Blum and Enrique Alba and Thomas Bartz-Beielstein and Daniele Loiacono and Francisco Luna and Joern Mehnen and Gabriela Ochoa and Mike Preuss and Emilia Tantar and Leonardo Vanneschi and Kent McClymont and Ed Keedwell and Emma Hart and Kevin Sim and Steven Gustafson and Ekaterina Vladislavleva and Anne Auger and Bernd Bischl and Dimo Brockhoff and Nikolaus Hansen and Olaf Mersmann and Petr Posik and Heike Trautmann and Muhammad Iqbal and Kamran Shafi and Ryan Urbanowicz and Stefan Wagner and Michael Affenzeller and David Walker and Richard Everson and Jonathan Fieldsend and Forrest Stonedahl and William Rand and Stephen L. Smith and Stefano Cagnoni and Robert M. Patton and Gisele L. Pappa and John Woodward and Jerry Swan and Krzysztof Krawiec and Alexandru-Adrian Tantar and Peter A. N. Bosman and Miguel Vega-Rodriguez and Jose M. Chaves-Gonzalez and David L. Gonzalez-Alvarez and Sergio Santander-Jimenez and Lee Spector and Maarten Keijzer and Kenneth Holladay and Tea Tusar and Boris Naujoks",
- isbn13 = "978-1-4503-1964-5",
- keywords = "genetic algorithms, genetic programming",
- pages = "1113--1120",
- month = "6-10 " # jul,
- organisation = "SIGEVO",
- address = "Amsterdam, The Netherlands",
-
DOI = "
doi:10.1145/2464576.2482690",
- publisher = "ACM",
- publisher_address = "New York, NY, USA",
-
abstract = "Historically, the quality of a solution in Genetic
Programming (GP) was often assessed based on its
performance on a given training sample. However, in
Machine Learning, we are more interested in achieving
reliable estimates of the quality of the evolving
individuals on unseen data. In this paper, we propose
to simulate the effect of unseen data during training
without actually using any additional data. We do this
by employing a technique called bootstrapping that
repeatedly re-samples with replacement from the
training data and helps estimate sensitivity of the
individual in question to small variations across these
re-sampled data sets. We minimise this sensitivity, as
measured by the Bootstrap Standard Error, together with
the training error, in an effort to evolve models that
generalise better to the unseen data.
We evaluate the proposed technique on four binary classification problems and compare with a standard GP approach. The results show that for the problems undertaken, the proposed method not only generalises significantly better than standard GP while the training performance improves, but also demonstrates a strong side effect of containing the tree sizes.",
- notes = "Also known as \cite{2482690} Distributed at GECCO-2013.",
