Generates candidates locally around promising regions, identified from
the archive. Good for exploitation/refinement.
Usage
candidate_generator_local(n_centers = 5L, radius = 0.1)
Arguments
- n_centers
(integer(1))
Number of archive points to sample around. Points are selected by
their y values (best first).
- radius
(numeric(1))
Sampling radius around each center, as a fraction of the domain range.
Default 0.1 (10% of range).
Value
A candidate generator function.
Details
Algorithm:
Select n_centers best points from archive
For each center, generate n/n_centers local samples
Local samples are uniform within a hypercube around the center
Note: This generator needs access to the archive, so the signature is
extended to function(search_space, n, archive).
