next up previous
Next: Symbolic Regression Up: Grammatical Evolution: A Steady Previous: Symbolic Integration


GE was successful in finding correct solutions to all of the problems described here. Results for the first two problems, that of Symbolic Regression and finding Trigonometric Identities, have been reported previously in [Ryan 98a] [Ryan 98b]. However, whilst carrying out the experiments for the Symbolic Integration problem it was decided to try a Steady State selection mechanism. The reasoning behind this was that GE produced a lot of invalid individuals, whom with a generational selection mechanism are passed onto the following generation, whereas with a Steady State mechanism these individuals would be less likely to survive into the next generation. Steady State selection was found to enhance the performance of the system and so, it was decided to apply this new selection mechanism to both of the problems previously tackled. The Steady State selection was also found to improve on the results reported previously for these problems and these results are reported here.