Tools
Sorted by:
Try your query at:
 |
 |
 |
 |
 |
 |
Results 11 - 20
of
92,103
Table I shows that the TBTP algorithm prohibits about 10% fewer turns than Up/Down. As expected, the simple spanning tree approach performs far worse, prohibiting about 90% of the turns in the network1. These results are almost insensitive to the network size. Figure 4 depicts the throughput performance with 99% confidence intervals. The results presented are in agreement with those obtained for the fraction of prohibited turns. We observe that, for 32 nodes, the TBTP achieves a throughput approximately 10% higher than that of Up/Down. Moreover, the relative difference in the performance between these two algorithms increases with the network size. Another important observation is that the throughput of TBTP is within a factor of at most 1.5 from that of the shortest-path scheme. This means that the cost of breaking all the cycles in the network
2004
Cited by 6
TABLE III COMPARISON OF EGO WITH PUBLISHED RESULTS OF RUNNING EANA [41] ON THE 11 LOWER-DIMENSIONAL FUNCTIONS. FOR EGO, WE RAN FOR EXACTLY a96a98a97a98a95 FUNCTION EVALUATIONS (COLUMN 2) AND QUOTE THE MEAN AND BEST COST FOUND FROM 51 RUNS (COLUMNS 3 AND 4, RESPECTIVELY). FOR EANA, THE ALGORITHM WAS STOPPED WHEN IT REACHED EITHER THE GLOBAL OPTIMUM (UP TO THE PRECISION OF THE COMPUTER) OR SOME UNSPECIFIED EVALUATION LIMIT [41]. THUS, THE MEAN NUMBER OF EVALUATIONS IS REPORTED (COLUMN 5) AND THE MEAN COST OF THE SOLUTION OVER THE 50 RUNS PERFORMED (LAST COLUMN). COMPARISON IS DIFFICULT BUT ON SOME FUNCTIONS, EGO OCCASIONALLY REACHES COMPARABLE FITNESS LEVELS ONE OR TWO ORDERS OF MAGNITUDE FASTER THAN THE MEAN NUMBER OF FUNCTION EVALUATIONS QUOTED FOR EANA. (EANA IS AN EVOLUTION STRATEGY, ALSO BASED ON BUILDING LANDSCAPE APPROXIMATIONS) Function EGO EANA
Table 4: Further tests on the QAP benchmark problems using the same pertur- bations and CPU times as before; given is the mean solution cost, averaged over 10 independent runs for each instance. Here we consider three different choices for the acceptance criterion. Clearly, the inclusion of diversification significantly lowers the mean cost found. instance acceptance 3 n/12 n/6 n/4 n/3 n/2 3n/4 n
2002
Cited by 56
Table 4: Mean costs of the GP classifier with and without the re-weighting method in comparison with C4.5
Table 23 Estimated mean cost of diagnosis, treatment and follow-up of colorectal cancer (with screening)
2004
Table 6 Reconstruction Budget and Interindustrial Economic Effects of Mean Cost Reconstruction Activity
"... In PAGE 15: ...nd workers residing outside the metropolitan area). In Table 5, leakages are $9.62 million. In Table6 , they are $29.67 million.... ..."
Table 4: Mean costs of the GP classifier with and without the re-weighting method in comparison with C4.5
Table 1 lists the results. The deviation from the lower bound is the same for both algorithms. This shows that the two- space genetic algorithm is able to find robust solutions.
1999
"... In PAGE 4: ...062 1.062 Table1 : Results for the Parallel Machine Scheduling Problem 6 Summary and Conclusions This paper presented a two-space genetic algorithm and sug- gested that it can be a general technique for solving minimax and robust discrete optimization problems. The two-space genetic algorithm should be useful for solving minimax op- timization problems in a wide variety of domains.... ..."
Cited by 9
Table 1 lists the results. The deviation from the lower bound is the same for both algorithms. This shows that the two- space genetic algorithm is able to find robust solutions.
"... In PAGE 4: ...062 1.062 Table1 : Results for the Parallel Machine Scheduling Problem 6 Summary and Conclusions This paper presented a two-space genetic algorithm and sug- gested that it can be a general technique for solving minimax and robust discrete optimization problems. The two-space genetic algorithm should be useful for solving minimax op- timization problems in a wide variety of domains.... ..."
Results 11 - 20
of
92,103
