This thesis describes the use of genetic programming to automate the discovery of numerical approximation formulae. Results are presented involving rediscovery of known approximations for Harmonic numbers and discovery of rational polynomial approximations for functions of one or more variables, the latter of which are compared to PadÃƒÂ© approximations obtained through a symbolic mathematics package. For functions of a single variable, it is shown that evolved solutions can be considered superior to PadÃƒÂ© approximations, which represent a powerful technique from numerical analysis, given certain tradeoffs between approximation cost and accuracy, while for functions of more than one variable, we are able to evolve rational polynomial approximations where no PadÃƒÂ© approximation can be computed. Furthermore, it is shown that evolved approximations can be iteratively improved through the evolution of approximations to their error function. Based on these results, we consider genetic programming to be a powerful and effective technique for the automated discovery of numerical approximation formulae.
Worcester Polytechnic Institute
All authors have granted to WPI a nonexclusive royalty-free license to distribute copies of the work. Copyright is held by the author or authors, with all rights reserved, unless otherwise noted. If you have any questions, please contact email@example.com.
Streeter, Matthew J., "Automated Discovery of Numerical Approximation Formulae Via Genetic Programming" (2001). Masters Theses (All Theses, All Years). 322.
genetic programming, approximations, machine learning, artificial intelligence, Approximation theory, Genetic programming (Computer science)