### Table 2. Search Space Properties

2004

"... In PAGE 9: ... Among the benchmarks there are four kernels: advect3d, lud, mm, vpenta and two full ap- plications: swim and mgrid. Table2 lists the applied transformations and the dimensions of the search space for each application. The total number of points... ..."

Cited by 8

### Table 6: Search space and word error rate #28WER#29 for 64k-word NAB task #28unigram LM look-

"... In PAGE 30: ... The acoustic pruning threshold was varied whereas both the LM pruning threshold and the maximum number of state hypotheses were kept #0Cxed. The results are shown in Table6 . These tests were carried out using a unigram LM look-ahead.... In PAGE 32: ...1#25. For the integrated search strategy using the trigram LM, Table 8 shows three recognition ex- periments that were selected from Table6 . These experiments result in search e#0Borts of 8 200, 15 800 and 27 000 state hypotheses and word error rates of 12.... ..."

### Table 1: Revised estimates of the search space for the first 32 codes

2006

"... In PAGE 3: .... equivalent case processed pruning (Section 6.4 of [2]). Table1 gives the revised estimates of the size of the search space for the first 32 codes. These 32 codes account for 91% of the search space.... ..."

### Table 2: Absolute Cost of Enumerating Various Search Spaces

2007

"... In PAGE 16: ... Before we explore that question, we will first compare the different search spaces directly. Table2 presents enumeration costs of various search spaces for stars and randomly generated queries. Enumeration al- gorithms are grouped by search space; the first row of each group shows the number of join operators in the search space5 (for randomly generated queries, we report the mean across 25 random instances).... In PAGE 17: ... [20] for the Columbia system (they report the number of join operators enumerated instead of CPU time, but similar patterns are evident). The disparity between search space sizes in Table2 re- inforces the efficacy of incremental optimization. Consider, however, an application demanding the globally optimal plan.... In PAGE 17: ... The third algorithm in each group performs exhaustive enumeration in both phases, il- lustrating the worst-case scenario where no pruning occurs so the first phase was completely wasted. The only timings in Table2 where a significant overhead is observed is for left-deep plans over star queries. In all other cases, the first phase adds only a small percentage to the overall cost.... ..."

### Table 4. Normalized Size of Search Space and Number of Evaluations Performed

2007

"... In PAGE 7: ...2 Search Space of PCPE Once we have established that PCPE actually obtains better results than PE and that the fitness value basically remains the same even when the proposed pruning techniques are applied, the next ques- tion we need to address is whether the pruning techniques proposed can actually reduce the search space to levels which are comparable to those of traditional PE. The upper half of Table4 shows the ratios of the number of configurations generated by the different PCPE algorithms versus those generated by traditional PE. Column Best PE shows the actual number of configurations generated by traditional PE.... In PAGE 7: ... This is an important advantage of this technique, since it copes directly with the main problem of PCPE. Another important conclusion which can be drawn from the upper half of Table4 is that, fortunately, the reduction of the search... In PAGE 8: ... Another important difference between PCPE and PE is that since the former allows computing several candidate residual pro- grams, we need to evaluate them in order to choose which is the best of them. The lower half of Table4 shows the number of con- figurations which need to be evaluated, i.e.... In PAGE 9: ... However, this will often be compensated by pruning, which reduces the number of configurations to be explored and of final solutions to be evaluated. All in all, we believe that the results shown in Table4 are indeed quite promising and show that the search space and the number of evaluations required by PCPE are manageable, when the proposed pruning techniques are applied. 7.... ..."

Cited by 1

### Table 1. The size of the search space for x86 instruction sequences of length 1 to 3. The last column shows the reduction in search space achieved through pruning and canonicalization.

in Processors

"... In PAGE 5: ... We currently check that all subsequences of length 2 are optimal using a table that lists all length 2 optimal sequences, when optimizing for codesize. Table1 lists the size of the set of enumerated instruction sequences with and without canonicalization and pruning. While canonical- ization provides the biggest reduction, the effect is cumulative and using both techniques we achieve over 50x improvement in the size of the search space for instruction sequences of length 3 on the x86 architecture.... In PAGE 5: ... While canonical- ization provides the biggest reduction, the effect is cumulative and using both techniques we achieve over 50x improvement in the size of the search space for instruction sequences of length 3 on the x86 architecture. In Table1 , the reduction due to pruning is only due to elimination of single instructions that are equivalent to other single instructions. For the codesize cost function, where we can employ the more aggressive pruning strategy, we get a total improvement of 60x (20% more) in the size of the search space at length 3.... ..."

### Table 3.5: The dynamic programming search algorithm with search space pruning.

2001

Cited by 4

### Table 3 Comparison of coding gains and decoding complexity (decoding search space dimension) for unitary DSTM for eight transmit antennas

"... In PAGE 4: ...5dB worse in BLER but has a much smaller decoding search space dimension of only 16 and a slightly higher spectral efficiency. Table3 compares the coding gain and decoding search space for three unitary DSTM schemes for eight transmit antennas. The first is our proposed DSTM based on O-STBC with spherical code.... ..."

### Table 1. Effect of range on search space Range ( a11 a1 ) 2-atomsets 3-atomsets Exec Time (s)

2002

"... In PAGE 6: ....1.2 Range Pruning Range pruning, as mentioned before, has a dramatic impact on reduction of search space. Table1 demonstrates this for the protein lysozyme for several different ranges. As ex- pected, search space, and thus run time, decreases as the range becomes more restrictive.... ..."

Cited by 12

### Table 2. Word accuracies obtained by the ASR system with dy- namic search-space pruning under different time constraints on Wall-Street-Journal. The first and second row show the perfor- mance of the system with preset-thresholds and without DSSP.

2002

Cited by 3