Geometric Generalisation of Surrogate Model-Based Optimisation to Combinatorial and Program Spaces
Table 2
Results for random search (RS), a genetic algorithm (GA), , and on QAP instances (kra30a, kra32, lipa30a, nug30, ste36a, and tho30) and unimodal instances (unih30 and unis30) of permutation problems.
Instance
RS
GA
Best
Average
SD
Best
Average
SD
Best
Average
SD
Best
Average
SD
kra30a
118730
122777.00
2034.86
115840
123445.80
2642.21
110850
119649.60
3389.70
117270
122027.00
2205.65
kra32
24156
25008.04
434.75
23440
24625.03
586.60
22590
24094.32
616.71
23848
24833.92
452.22
lipa30a
13664
13710.08
16.38
13646
13704.42
22.53
13633
13700.52
21.89
13638
13696.32
22.03
nug30
7350
7618.00
84.60
7296
7558.36
126.65
7276
7500.24
97.36
7328
7563.88
94.38
ste36a
16736
18335.56
602.04
16516
18287.96
841.59
15364
17311.00
922.20
15654
17840.68
889.74
tho30
190256
196662.82
3211.20
180274
194389.91
4414.22
180860
193415.44
4629.03
186172
195231.92
3534.21
unih30
24
25.80
0.78
22
25.04
1.26
17
20.84
1.55
21
25.18
0.95
unis30
19
21.66
0.87
17
20.91
1.29
15
18.48
1.83
19
21.04
0.94
The best, average, and standard deviation of the best fitness found by each algorithm are reported for 50 runs.
