Tools
Sorted by:
Try your query at:
 |
 |
 |
 |
 |
 |
Results 1 - 10
of
12,508
Table 6: Closed gap (%)
"... In PAGE 14: ...t is not complete. Computations are in progress. Table 5: Computational results on hard instances others. By analyzing the results of Table6 , where the closed gap of exact separation is presented for the all instances, we can observe that exact separation does not significantly improve the lower bound at the root node for the unsolved instances and this calls for further polyhedral investigation of the GAP polytope. 5 Conclusions This paper reports on the computational experience with an exact knapsack separation procedure for the Generalized Assignment Problem.... ..."
Table A.2: shows the issues raised in the cognitive dimension (CD) questionnaire for participant. This are the original results before the reclassification. The issues raised are listed in Table 4.5. The cognitive dimensions are 1. visibility and juxtaposability, 2. viscosity, 3. hard mental operation, 4. error proneness, 5. closeness of mapping, 6. role expressiveness and 7. consistency.
1991
Cited by 3
Table 3: The complexity of Schur functions on 1-semi-unitary matrices This analysis seems to suggest a hierarchy of di culty: The determinant is always easy. For a given class of matrices M, whenever the permanent is di cult, so are ham and cycle . Finally, whenever cycle? is hard, so is cycle+. Note that the unitary matrices and C( x; y) are closed under some kind of multiplication and the C( x; y) matrices are themselves unitary. Problem 5.1 Is this hierachy of complexity of Schur functions true for any multi- plicatively closed class of matrices ?
in 9
Table 2 provides detailed information about the parallel runs. All speedup values were computed relative to the respective sequential applications, with the Satin-specific annotations removed from the code. Both TSP and IDA* achieve close to linear speedup, knapsack achieves lower performance, whereas Fibonacci, as expected, has hardly any speedup.
2000
"... In PAGE 6: ...Figure 3: Application speedups There is a strong correlation between measured speedup and the sequential overhead value, as already shown in Table 1: the lower the overhead, the higher the speedup we achieved. In Table2 we compare the measured speedup with its upper bound, computed as the number of CPUs divided by the overhead on a sin- gle CPU. We also show the relative speedup, the percent- age of this upper bound as actually achieved by the mea- sured speedup.... In PAGE 6: ... The actual percent- age depends (like the sequential overhead) on the number of method parameters and their total size. Table2 also lists the total number of spawned jobs and the number of stolen jobs, which is less than 1 out of 5000 for all appli- cations. Because the number of stolen jobs is so small, speedups are mainly determined by sequential overhead.... In PAGE 7: ...Table2 : Parallel performance breakdown for 32 CPUs application overhead speedup #CPUs/overhead relative speedup jobs stolen fib 11.90 2.... ..."
Cited by 10
Table 1. Different complexity measures for k-way distribution and k-way merging in comparison based sorting algorithms. All these algorithms need n log k element com- parisons. Lower order terms are omitted. Branches: number of hard to predict branches; data dep.: number of instructions that depend on another close-by instruction; I/Os: number of cache faults assuming the I/O model with block size B; instructions: neces- sary size of instruction cache.
"... In PAGE 5: ... One goal of this paper is to give an example of how it can be algorithmically interesting and relevant for performance to consider complexity measures beyond the RAM model and memory hierarchies. Table1 summarizes the complexity of k-way distribution and k-way merging for five different algorithms and five different complexity measures (six if we count comparisons). For the sake of this comparison, log k recursion levels of quicksort are viewed as a means of k-way distribution.... ..."
Table 1: Constraint table for CD recorder.
"... In PAGE 4: ...Table1 . In the constraint table, the single shaded cell in each row may be thought of as the if part of a set of if - then statements, with the other cells in the same row giving the then parts.... ..."
Table 1: Constraint table for CD recorder.
"... In PAGE 4: ...Table1 . In the constraint table, the single shaded cell in each row may be thought of as the if part of a set of if - then statements, with the other cells in the same row giving the then parts.... ..."
Table 4: Contingency table: Redundant Assignments vs. Hard Bugs. There are 345 les with both error types, 435 les with an assign error and no hard bugs, 206 les with a hard bug and no assignment error, and 1069 les with no bugs of either type. A T-statistic value above four gives a p-value of less than .05, which strongly suggests the two events are not independent. The observed T value of 194.37 gives a p-value of essentially 0, noticeably better than this standard threshold. Intuitively, the correlation between error types can be seen in that the ratio of 345/435 is considerably larger than the ratio 206/1069 | if the events were independent, we expect these two ratios to be close.
Table 11: Computational results for hard instances of 1jrjj P wjCj.
2002
"... In PAGE 10: ... We note that the duality gap is larger than for the Optimal instances. This can be easily explained by observing ( Table11 ) that for these instances the quality of the lower bound yielded from the time-indexed formulation, reported in the column LP-LB, is poor. Nevertheless the lagrangian heuristic confirms to be robust, since the upper bound is very close (less than 1%) to the value of the optimal solution computed by Cplex 7:0, reported in the column Opt.... ..."
Table 2: HARD Track, hard relevance evaluation
2004
Cited by 15
Results 1 - 10
of
12,508
