Returns all local optimums of the acquisition function, no matter the utility.
Usage
getLocalOptimums(
optObj,
bounds = optObj$bounds,
acq = optObj$optPars$acq,
kappa = optObj$optPars$kappa,
eps = optObj$optPars$eps,
convThresh = optObj$optPars$convThresh,
gsPoints = optObj$optPars$gsPoints,
parallel = FALSE,
verbose = 1
)Arguments
- optObj
an object of class
bayesOpt. The following parameters are all defaulted to the options provided in this object, but can be manually specified.- bounds
Same as in
bayesOpt()- acq
Same as in
bayesOpt()- kappa
Same as in
bayesOpt()- eps
Same as in
bayesOpt()- convThresh
Same as in
bayesOpt()- gsPoints
Same as in
bayesOpt()- parallel
Same as in
bayesOpt()- verbose
Should warnings be shown before results are returned prematurely?
Value
A data table of local optimums, including the utility (gpUtility), the utility relative to the max utility (relUtility), and the steps taken in the L-BFGS-B method (gradCount).
Details
gsPoints points in the parameter space are randomly initialized, and
the L-BFGS-B method is used to find the closest local optimum to each point.
dbscan is then used to cluster points together which converged to the same
optimum - only unique optimums are returned.
Examples
scoringFunction <- function(x) {
a <- exp(-(2-x)^2)*1.5
b <- exp(-(4-x)^2)*2
c <- exp(-(6-x)^2)*1
return(list(Score = a+b+c))
}
bounds <- list(x = c(0,8))
Results <- bayesOpt(
FUN = scoringFunction
, bounds = bounds
, initPoints = 3
, iters.n = 2
, gsPoints = 10
)
#> Ran initial scoring function 3 times in 1 thread: 0.684999999999999 seconds.
#> Starting Epoch 1
#> 1) Fitting Gaussian Process...
#> - Convergence Not Found. Trying again with tighter parameters...
#> - Convergence Not Found. Trying again with tighter parameters...
#> - Convergence Not Found. Trying again with tighter parameters...
#> - Maximum convergence attempts exceeded - process is probably sampling random
#> points.
#> 2) Ran local optimum search: 1.018 seconds.
#> 3) Ran `FUN()` 1 time in 1 thread: 0.234999999999999 seconds.
#> Starting Epoch 2
#> 1) Fitting Gaussian Process...
#> 2) Ran local optimum search: 0.0499999999999972 seconds.
#> 3) Ran `FUN()` 1 time in 1 thread: 0.233000000000004 seconds.
print(getLocalOptimums(Results))
#> localOptimum x gpUtility relUtility gradCount
#> <int> <num> <num> <num> <num>
#> 1: 1 8.000000 0.5330484 0.9787271 2
#> 2: 2 4.558688 0.5446344 1.0000000 7
#> 3: 3 1.481529 0.3684267 0.6764661 9