As can be seen from the results in Table 4, Steady State selection was responsible for the improved performance in all three problems dealt with here. The use of a Steady State selection mechanism, especially with larger populations [De Jong 92], has been regarded as a way to improve performance of a Genetic Algorithm. In our case the fact that so many invalid individuals were being produced when creating the next generation, caused the system to perform poorly, because these invalid individuals were being passed onto the next generation as a result of the generational selection mechanism. However with Steady State selection this problem is overcome by only replacing a few individuals at every new generation, thus preventing these invlaid individuals from entering the population.
Another possible reason for the improvement in performance could also be attributed to the fact that using Steady State selection means those individuals with a better fitness value are more likely to be preserved from one generation to the next, however poorer individuals are more prone to deletion from the population. It is likely that individuals (or parts of) that represent useful building blocks are more likely to be preserved when using a Steady State as opposed to a Generational selection mechanism. In effect the population as a whole is working toward the solution.
Comparing Figures 5 and 6, it can be seen that the average fitness with the Steady State selection in Figure 5 is greater than that in Figure 6, especially as the run nears completion. This supports the theory that the population is working toward the solution as a whole when using Steady State.