next up previous contents
Next: Activity Measures Up: Detailed Analysis of a Previous: Population Size and Diversity

Flaw Period and Proportion of Unfaithful Replications

The two measures used to track evolution of the programs along the A-fidelity axis were: the flaw period of individual programs, and the proportion of the total number of offspring produced by an individual that were unfaithful (less than 100% accurate). The corresponding graphs are shown in Figures 5.6 and 5.11 respectively. In the graph of unfaithful replications (Figure 5.11), the vast majority of the points lie at zero, indicating that nearly all programs which reproduced did so faithfully. In order to see whether there was any trend for programs which did not always reproduce faithfully, the data from Figure 5.11 is reproduced in Figure 5.12 with the exception that points lying at zero infidelity have been omitted. The bar on the right of these two figures shows the scale (i.e. the mapping between the darkness of the plot at any point in the graph and the number of programs which had that particular infidelity at that particular time). Note that the scale in Figure 5.12 is two orders of magnitude smaller than in Figure 5.11.


  
Figure: Standard Run: Unfaithful Replications.
\resizebox{!}{0.75\linewidth}{\includegraphics{graphs/standard/punfaithfulStandard.shrunk.ps}}





Figure: Standard Run: Unfaithful Replications (omitting points with zero infidelity).
\resizebox{!}{0.75\linewidth}{\includegraphics{graphs/standard/punfaithfulStandard_crop.shrunk.ps}}



It is clear from these graphs that there was little change in either of these measures throughout the run. In other words, the programs are not evolving along the A-fidelity axis. From the analysis in the previous sections, it would appear that the parameter choices for this run create a selection pressure that predominantly favours evolution along the longevity and fecundity axes.


next up previous contents
Next: Activity Measures Up: Detailed Analysis of a Previous: Population Size and Diversity
Tim Taylor
1999-05-29