Master's Thesis

Master of Computer Science Carleton University

Thesis Title: Program Induction Using an Unmodified Genetic Algorithm.
Supervisor: Professor Franz Oppacher
University: Carleton University, Ottawa, Canada
Call Number: Carleton University - M.C.S. 1993.W55
Conferred: September 1993

Full version: PDF version (coming soon)


My thesis was an attempt to study Genetic Programming (GP) in relation to the Genetic Algorithm (GA). Since GP chromosomes are non-homologous and use a special purpose crossover (and do not require the use of mutation), it is not clear that the theoretical foundations developed for the GA hold for GP. In the thesis, an algorithm (called EPI for Evolutionary Program Induction) was developed that stayed within the canonical Genetic Algorithm paradigm yet bred programs in a similar manner to GP. EPI was tested on three problems that had known behavior under GP; EPI performed identically to GP (no statistical significance was detected) over this test suite. The success of the implementation showed that the modified crossover used by GP is not necessary to solve program induction when using a GA.