The material printed in this thesis has not been previously submitted for a degree or diploma in any university, and to the best of my knowledge contains no material previously published or written by another person except where due acknowledgement is made in the thesis itself.
 

Mr Colin Thorne:
My Supervisor, whose perspective, direction and observations were insightful and objective. In particular, for the confidence to afford me a comfortable level of autonomy throughout this research. I hope this confidence was not misgiven.

Dr Clyde Wild:
For his help, expertise and enthusiasm beyond the call of duty.

Mr Peter J. Hackett:
My father, a pedant? Perhaps, but whose interest, encouragement and efforts were invaluable to the preparation of this paper.

Mr Hans Grahlmann:
Whose personal style of encouragement was instrumental in my decision to take this honours year on. Also, for his efforts to afford me another year, free of distraction.
 
 

  Genetic programming is applied to a benchmark version of the cart-pole problem. The effect of three selection techniques (roulette wheel, expected value model and tournament selection) are investigated. The resultant on-line and off-line learning performances are compared.The two stochastic selection techniques (roulette wheel and expected value model) are found to outperform tournament selection (a competitive strategy) at the on-line learning of balancing the pole and centring the cart from a difficult starting position. For off-line learning, no significant difference is found between the three selection strategies.