RGA - algorithm for black box global optimization
![]()
Fig.3. Testing of RGA with one of standart test-functions:
gradient has jumps along a few lines but vanishes in a broad vicinity of minima
[ Sen, M. K., and P. L. Stoffa, Global Optimization Methods in Geophysical Inversion,
as part of Advances in Exploration Geophysics Series, Elsevier Publishing Co.,
The Netherlands, 1995 ].
The sequence of images is similar to that on Fig.2:
a) the true OF in the range [-10.:10.]x[-10.:10.] with global minimum 0. at the point (0., 0.);
b) the 1st movie is composed with 9 sequential tries,
c) the 2nd one represents the first zoom,
d) the final 3rd movie shows the second zoom before RGA -auto-termination.The global minimum is detected at the point (0.0067, 0.0600) with the value 0.00000037. The total number of OF calls is 541.NB! Please, pay attention: isolines are not equidistant, but strongly dense near '0' and '1'.
a)
b)
c)
d)
NB! All images of this page are gathered in a multipaged PDF-file
(~1.9 Mb)
..
.