Evolving potentiallv-infinite list comprehensions
Lee Braine, Rüdi Lutz · World Automation Congress · 2004
This paper introduces potentially-infinite list comprehensions into evolutionary computation. List comprehensions are programming constructs based on Zermelo-Fraenkel (ZF) set theory and are used in modern functional languages to define sets concisely. We present n new, higher-order, polymorphic and strongly-typed Genetic Programming (GP) system. ZF-GP, that evolves potentially-infinite list comprehensions. The resulting language has a highly-focussed search space that is smaller than is typical in evolutionary computation. Experiments demonstrate that the ZF-GP system can evolve list comprehension solutions to formula-fitting problems more efficiently than random search.