"... In PAGE 13: ... Table2 . Circuit testing of fault testing for the circuits of Table 1.... In PAGE 13: ...i.e. every fault made F satis- flable). The second column of Table2 gives the size of a test set. The third column gives the result for a test set detecting all stuck-at faults in N.... In PAGE 13: ...The value in parentheses shows the worst result out of 10. For example, for the circuit c432, in the flrst experiment (flrst line of Table2 ) we generated 10 test sets, each consisting of 58 tests. On average, 69.... In PAGE 13: ...7 and the worst result was 76 faults. One can draw at least three conclusions from Table2 . First, the quality of a test set extracted from a resolution proof depends on proof quality.... In PAGE 14: ...Table2 shows that these tests performed better than tests extracted from proofs found by FI (that are signiflcantly larger). Second, even though tests detecting the stuck-at faults have high quality, they do not detect all literal appearance faults.... ..."

"... In PAGE 7: ... In the second scenario, we use a lower-end machine (SGI Indigo2 Extreme connected with a 100- BaseT switched Ethernet) for the rendering phase and the Reality Monster for the isosurface extraction. The results for the first scenario, Table1 , shows that for larger and more complex isosurfaces the new method still outperforms extraction of the full isosur- face, though the performance falls short when only a change of the view point is considered. The results for the second scenario, Table 2, demon- strate the advantage of this method.... ..."

"... In PAGE 7: ... The results for the first scenario, Table 1, shows that for larger and more complex isosurfaces the new method still outperforms extraction of the full isosur- face, though the performance falls short when only a change of the view point is considered. The results for the second scenario, Table2 , demon- strate the advantage of this method. Even when the time for the extraction is added to each new view po- sition, the new method out performs a regular full oc- tree traversal and extraction due to the LAN bandwidth limitations.... ..."

"... In PAGE 7: ... In the second scenario, we use a lower-end machine (SGI Indigo2 Extreme connected with a 100- BaseT switched Ethernet) for the rendering phase and the Reality Monster for the isosurface extraction. The results for the first scenario, Table1 , shows that for larger and more complex isosurfaces the new method still outperforms extraction of the full isosur- face, though the performance falls short when only a change of the view point is considered. The results for the second scenario, Table 2, demon- strate the advantage of this method.... ..."

"... In PAGE 7: ... The results for the first scenario, Table 1, shows that for larger and more complex isosurfaces the new method still outperforms extraction of the full isosur- face, though the performance falls short when only a change of the view point is considered. The results for the second scenario, Table2 , demon- strate the advantage of this method. Even when the time for the extraction is added to each new view po- sition, the new method out performs a regular full oc- tree traversal and extraction due to the LAN bandwidth limitations.... ..."

"... In PAGE 9: ...5 Results for the nine Art Styles In this section, we present the results for all nine art styles for which the relation instances were extracted. In Table3 , we show the precision and the number of correct relation instances extracted for each of the nine art styles for an arbitrarily chosen value for the two threshold parameters (DF=0.3 and Max=20).... ..."

"... In PAGE 9: ...000 13 Surrealist 0.714 5 In Table4 , we list both the average precision and the total sum of the number of correct relation instances extracted for the nine art styles for 24 combinations of the two threshold parameters DF and Max. The lowest value for precision is 0.... ..."