Large-Scale Evolutionary Strategy Based on Gradient Approximation
Table 12
CPU computational time of improved CMA-ES and non-improved CMA-ES.
D = 1000
F1
F2
F3
F4
F5
F6
F7
F8
F9
F10
MA-SW-Chain
166.87
218.9
197.29
199.92
260.2
229.17
169.57
165.96
431.78
478.9
CMA-ES
159.14
216.73
184.41
141.91
232.9
237.31
236.01
136.01
186.61
198.43
GI-ES
147.96
213.04
180.72
98.22
189.21
193.62
192.32
92.32
142.92
154.74
D = 1000
F11
F12
F13
F14
F15
F16
F17
F18
F19
F20
MA-SW-Chain
472.41
169.31
169.96
687.67
735.8
725.51
174.55
180.11
163.41
174.07
CMA-ES
308.27
106.25
137.12
233.54
372.58
277.15
36.5
38.39
136.57
138.22
GI-ES
251.96
92.54
93.41
189.85
328.89
333.46
92.81
94.7
192.88
194.53
MA-SW-Chain is also used as a comparison. The table records the average time of 51 independent runs of the three algorithms when the dimension is 1000. The maximum number of iterations is 10E06.