next up previous
Next: 8.2 Up: 8. The Influence of Previous: 8. The Influence of

8.1 $E_1 \vert E$

The first example reveals two characteristics. We have that for all orderings computed the strategy I-R is better than NONE, followed by I-ALL and P-R. Further, the strategies P-ALL, P-G, P-R and I-R-P, that is all strategies involving prefixes, are equally bad. On the other hand, for all strategies the syllable orderings perform better than the Knuth-Bendix and the length-lexicographic orderings. The performance of the latter is rather similar.

Further, this example shows that the differences between different orderings for the same strategy can be quite large. If we consider the best strategy found using a length-lexicographical ordering, namely I-R, we see that the number of cosets defined ranges from 59/89 to 97/97 (maximal/total). Remarkably, the best ordering is syl-r-tsrTSR while ll-tsrTSR is the one performing worst. For I-ALL the values range from 99/99 for syl-l-tsrTSR to 547/547 for ll-tsrTSR. Table 12 lists the orderings while Table 13 lists the results.


next up previous
Next: 8.2 Up: 8. The Influence of Previous: 8. The Influence of
| ZCA Home | Reports |