Global optimization of blackbox functions by iteratively shrinking the parameter space around good points.
In each iteration points
random points are drawn from the full param space, and the best point is selected
for shrinkage.
Now supports trafo functions in the ParamSet
.
# Search through a fully numeric param space
f = smoof::makeSphereFunction(2)
# Create a (vectorized) blackbox optimisation function
fn = function(x) apply(x, 1, f)
# Define focussearch parameters
ctrl = makeFocusSearchControl(maxit = 5, restarts = 3, points = 100)
# Define searchable parameter space
ps = makeParamSet(
makeNumericParam("x1", lower = 0, upper = 10),
makeNumericParam("x2", lower = 0, upper = 10)
)
# Run focussearch
focussearch(fn, ps, ctrl)