Table 2 Results of the experiments. For each algorithm, quantiles Q25, Q50, Q75 and Q90 for the number of trials for simple test classes are presented.

From: On the efficiency of nature-inspired metaheuristics in expensive global optimization with limited budget

N

Q

Metaheuristic algorithms (10,000 runs for each algorithm and class)

Deterministic algorithms (100 runs for each algorithm and class)

Differential Evolution

Particle Swarm Optimization

Genetic Algorithm

Artificial Bee Colony

Firefly Algorithm

DIRECT

DIRECT-L

ADC

2

25

0.07%

0.04%

0.13%

0.03%

0.05%

0.01%

0.01%

0.01%

50

0.14%

0.08%

3.28%

0.05%

0.08%

0.01%

0.02%

0.02%

75

0.26%

0.21%

99.53%

0.12%

0.14%

0.02%

0.04%

0.02%

90

0.60%

95.74%

100.00%

0.69%

0.22%

0.04%

0.07%

0.02%

3

25

0.22%

0.29%

1.35%

0.09%

0.43%

0.02%

0.03%

0.05%

50

0.44%

0.92%

4.83%

0.16%

0.84%

0.04%

0.06%

0.06%

75

1.99%

3.60%

26.64%

1.06%

1.66%

0.15%

0.19%

0.10%

90

35.70%

97.61%

99.83%

2.94%

2.91%

0.30%

0.57%

0.12%

4

25

0.17%

0.25%

0.24%

0.12%

0.32%

0.13%

0.23%

0.28%

50

2.49%

1.04%

0.83%

1.34%

0.83%

0.50%

0.73%

0.41%

75

89.48%

97.25%

3.78%

6.03%

2.69%

1.12%

3.07%

0.73%

90

100.00%

100.00%

96.73%

15.89%

6.18%

2.21%

5.57%

0.89%

5

25

0.17%

0.27%

0.46%

0.12%

0.33%

0.06%

0.32%

0.21%

50

0.29%

0.87%

1.56%

0.34%

0.62%

0.16%

0.93%

0.39%

75

97.05%

4.04%

9.24%

3.73%

2.86%

0.64%

1.90%

0.64%

90

100.00%

99.78%

100.00%

10.89%

14.11%

1.54%

3.54%

1.02%