A genetic programming system for the induction of iterative solution algorithms to novice procedural programming problems

摘要

The study presented in this paper evaluates genetic programming (GP) as a means of evolving solution algorithms to novice iterative programming problems. This research forms part of a study aimed at reducing the costs associated with developing intelligent programming tutors by inducing solutions to the programming problems presented to students, instead of requiring the lecturer to provide these solutions. The paper proposes a GP system for the induction of algorithms using iteration and nested iteration. The proposed system was tested on 15 randomly selected novice procedural programming problems requiring the use of iterative and nested-iterative constructs. The system was able to evolve solutions to eight of these problems. Premature convergence of the GP algorithm as a result of fitness function biases was identified as the cause of the failure of the system to induce solutions to the remaining seven problems. The iterative structure-based algorithm (ISBA) was developed and successfully implemented to escape local optima caused by fitness function biases and evolve solutions to these problems.