### Table 1.8. Randomized rapid restarts (RRR) versus deterministic versions of backtrack search procedures (Satz solver used on SAT encodings; Ilog solver on CSP encodings).

### Table IV. Randomized rapid restarts (RRR) versus determinis- tic versions of backtrack search procedures (Satz solver used on SAT encodings; Ilog solver on CSP encodings).

2000

Cited by 91

### Table IV. Randomized rapid restarts (RRR) versus determin- istic versions of backtrack search procedures (Satz solver used on SAT encodings; Ilog solver on CSP encodings).

2000

Cited by 91

### Table 1 Comparison of the H2SOM to similar sized standard SOMs. The table shows the training times in hours and minutes for the map formation of the 60000 training samples and the seconds for the best match lookups for the 10000 test samples of the MNIST database. For the H2SOM the test runs were performed with (a) the rapid SF-search with k = 2 and (b) a slower global search. (All results were obtained on a standard laptop with 1.5 GHz Pentium-M processor). SOM H2SOM, nb = 8

"... In PAGE 9: ... To each test item then the class label of its corresponding best match node is assigned. In Table1 the classification accuracies for different SOMs are given. Again, the most prominent difference is the time needed for the training of the networks.... ..."

### Table 1 summarizes the predictive accuracies and learning times of the randomized rapid restarts technique vs. the standard exhaustive breadth- first top-down search algorithm. The former method was tested with the Acc parameter set to the values 0.7 and 0.9. Similarly, the latter method was tested with two values 0.7 and 0.9 of the minimum accuracy require- ment on a clause to be accepted for the constructed theory. The results suggest that by using randomized rapid restarts we achieved a drastic reduction of the search times for the price of only a small loss in predictive accuracy.

2003

"... In PAGE 12: ... Table1 . Estimated predictive accuracies (A) and theory construction times (T).... ..."

Cited by 21

### Table 3.2: Path Expansion in A* Search of Figure 2.4, these estimates are shown in Figure 3.3. Consider the expansion of the path f(B0; S0); (B1; S1)g in Figure 3.4. The esti- mates generated from the Viterbi search are given in parentheses at the destination nodes. In Table 3.2, the left side shows the basic stack decoder expansion, and the right side shows the inclusion of the estimate. Note that the insertion to (B2; S1), which is along the best path, is ranked last without the estimate but rst with the estimate. The estimate helps the algorithm overlook the match with cost 14 in favor of the insertion with cost {10. With Viterbi estimates, the A* search algorithm will rapidly reconstruct the best path. Unfortunately, the backwards Viterbi search cannot be initiated until the comple- tion of the utterance. A forward Viterbi search followed by a backward A* search

1992

Cited by 4

### Table 1. The size of search space for searching the para- meters for membership function for 256 gray level image

1998

"... In PAGE 3: ... SIMULATED ANNEALING ALGORITHM The problem mentioned in Section 3 can be con- sidered as a combinatorial optimization problem. We use Table1 to explain why we need a searching algorithm rather than an exhausted search to nd a solution in the search space. In Table 1, we can see that the size of search space increases very rapidly when the number of parameters increases.... In PAGE 3: ... We use Table 1 to explain why we need a searching algorithm rather than an exhausted search to nd a solution in the search space. In Table1 , we can see that the size of search space increases very rapidly when the number of parameters increases. In the case mentioned in the previous sec- tion, the search space is about 2.... ..."

Cited by 2

### Table 6: Iterations needed for the modifled problem instances from Solomon

in A

"... In PAGE 15: ... The set of feasible solutions to VRPTW becomes very restricted as the ratio grows, resulting in a rapid search. Table6 shows the average number of iterations needed to run a test problem of various classes. As depicted in Tables 5 and 6, the iteration number decreased as the demand ratio increased.... ..."

### Table IV also lists the time to build the complete search structure for six-way search for each of the four databases. We made no attempt to optimize the table-building process; it seems very likely that the numbers can be improved. From the table, we see that the cost of building a complete table can range from 5.8 s (for Mae-East) to 350 ms (for Paix). At first glance, this seems to preclude the use of binary search for large backbone routers because of the potential need for rapid insertion times (10-100 ms). However, incremental insertion times can be much smaller, even for large databases like Mae-East.

1998

Cited by 71