On sampling error in genetic programming

Created by W.Langdon from gp-bibliography.bib Revision:1.8051

@Article{Schweim:NatComput,

• author = "Dirk Schweim and David Wittenberg and Franz Rothlauf",
• title = "On sampling error in genetic programming",
• journal = "Natural Computing",
• year = "2022",
• volume = "21",
• pages = "173--186",
• keywords = "genetic algorithms, genetic programming, Sampling error, Initial supply, Building blocks, Initial population, Ramped half-and-half, Full, Grow, n-Grams",
• ISSN = "1567-7818",
• URL = "https://rdcu.be/cmvW6",
• DOI = "doi:10.1007/s11047-020-09828-w",
• size = "14 pages",
• abstract = "The initial population in genetic programming (GP) should form a representative sample of all possible solutions (the search space). While large populations accurately approximate the distribution of possible solutions, small populations tend to incorporate a sampling error. This paper analyzes how the size of a GP population affects the sampling error and contributes to answering the question of how to size initial GP populations. First, we present a probabilistic model of the expected number of subtrees for GP populations initialized with full, grow, or ramped half-and-half. Second, based on our frequency model, we present a model that estimates the sampling error for a given GP population size. We validate our models empirically and show that, compared to smaller population sizes, our recommended population sizes largely reduce the sampling error of measured fitness values. Increasing the population sizes even more, however, does not considerably reduce the sampling error of fitness values. Last, we recommend population sizes for some widely used benchmark problem instances that result in a low sampling error. A low sampling error at initialization is necessary (but not sufficient) for a reliable search since lowering the sampling error means that the overall random variations in a random sample are reduced. Our results indicate that sampling error is a severe problem for GP, making large initial population sizes necessary to obtain a low sampling error. Our model allows practitioners of GP to determine a minimum initial population size so that the sampling error is lower than a threshold, given a confidence level.",
• notes = "Johannes Gutenberg University, Jakob-Welder-Weg 9,55128 Mainz, Germany",

}

Genetic Programming entries for Dirk Schweim David Wittenberg Franz Rothlauf

Citations