Description Usage Arguments Details Value References See Also
Hierarchical test functions by Horn and Zaefferer
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 | makeHZHF(id, in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF01(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF02(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF03(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF04(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF05(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF06(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF07(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF08(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF09(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF10(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF11(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF12(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF13(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF14(in.dim, k, c = 0.5, s = 0, check = TRUE)
makeHZHF15(in.dim, k, c = 0.5, s = 0, check = TRUE)
|
id |
[ |
in.dim |
[ |
k |
[ |
c |
[ |
s |
[ |
check |
[ |
Hierarchical test functions have several parameters that only influence the outcome if a certain condition is fullfilled. The HZHF test function family consists of 15 hierarchical test functions with numeric parameters z[1], ... z[in.dim]. The first k of these parameters always affect the function value while the last in.dim - k ones only do if z[1] fullfills a certain condition.
All HZHF functions are to be minimized and follow the scheme
f(z) = a(z[1], ..., z[k]) + if (z[1] < c) b(z[1], ..., z[in.dim]) else s
Here, c controls the size of the active region of the hierarchical parameters. s is a shift added to the function if the hierarchical parameters are inactive and controls a) whether an imputation approach is promising and b) the jump height at the discontinuity.
The function a and b are defined using different transformation of the WFG toolkit (A toolkit for multi-objective test functions). The 15 HZHF function differ in their properties, as for example modality and seperability.
For s = 0 and c < 0.7 the global optimum is allways at z[i] = 0.7 * i, i = 1, ..., in.dim (the hierarchical parameters are active). If s = 0 and c > 0.7 holds, the optimum is at z[i] = 0.7 * i, i = 1, ..., k, z[i] = NA, i = k + 1, ..., in.dim, i.e. the hierarchical parameters are inactive. The best possible function value is 0 in both cases.
If s < 0 and the optimum is in in the inactive region, its function value is shifted to s, but its location remains unchanged. If s < 0 and the optimum is in the active region, the former global optimum is only guarenteed to be a local optimum with function value 0. Their may be other local optima in the inactive regions with function values < 0, especially at the discontinuity z[1] = c. Unfortunately, we can not control those values.
If s > 0 and the optimum is in the active region, it is guarenteed to be the global optimum with function value 0. If, however, s > 0 and the optimum is in the inactive region, it is only guarenteed to be a local optimum with function value s. There may be other local optima in the active regions with function values between 0 and b, especially at the discontinuity z[1] = c. Unfortunately, we can not control those values.
A hzhf_function
, subclass of smoof_function
.
Huband, Simon ; Hingston, Phil ; Barone, Luigi ; While, Lyndon: A Review of Multiobjective Test Problems and a Scalable Test Problem Toolkit. In: IEEE Trans. on Evolutionary Computation 10 (2006), No. 5, pp. 477-506
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.