Single-Objective Bayesian Optimization via Multipoint Constant Liar

Loop function for single-objective Bayesian Optimization via multipoint constant liar. Normally used inside an OptimizerMbo.

In each iteration after the initial design, the surrogate and acquisition function are updated. The acquisition function is then optimized, to find a candidate but instead of evaluating this candidate, the objective function value is obtained by applying the liar function to all previously obtained objective function values. This is repeated q - 1 times to obtain a total of q candidates that are then evaluated in a single batch.

Usage

bayesopt_mpcl(
  instance,
  surrogate,
  acq_function,
  acq_optimizer,
  init_design_size = NULL,
  q = 2L,
  liar = mean,
  random_interleave_iter = 0L
)

Arguments

instance: (bbotk::OptimInstanceBatchSingleCrit)
The bbotk::OptimInstanceBatchSingleCrit to be optimized.
surrogate: (Surrogate)
Surrogate to be used as a surrogate. Typically a SurrogateLearner.
acq_function: (AcqFunction)
AcqFunction to be used as acquisition function.
acq_optimizer: (AcqOptimizer)
AcqOptimizer to be used as acquisition function optimizer.
init_design_size: (NULL | integer(1))
Size of the initial design. If NULL and the bbotk::ArchiveBatch contains no evaluations, 4 * d is used with d being the dimensionality of the search space. Points are generated via a Sobol sequence.
q: (integer(1))
Batch size > 1. Default is 2.
liar: (function)
Any function accepting a numeric vector as input and returning a single numeric output. Default is mean. Other sensible functions include min (or max, depending on the optimization direction).
random_interleave_iter: (integer(1))
Every random_interleave_iter iteration (starting after the initial design), a point is sampled uniformly at random and evaluated (instead of a model based proposal). For example, if random_interleave_iter = 2, random interleaving is performed in the second, fourth, sixth, ... iteration. Default is 0, i.e., no random interleaving is performed at all.

Value

invisible(instance)
The original instance is modified in-place and returned invisible.

Note

The acq_function$surrogate, even if already populated, will always be overwritten by the surrogate.
The acq_optimizer$acq_function, even if already populated, will always be overwritten by acq_function.
The surrogate$archive, even if already populated, will always be overwritten by the bbotk::ArchiveBatch of the bbotk::OptimInstanceBatchSingleCrit.
To make use of parallel evaluations in the case of `q > 1, the objective function of the bbotk::OptimInstanceBatchSingleCrit must be implemented accordingly.

References

Ginsbourger, David, Le Riche, Rodolphe, Carraro, Laurent (2008). “A Multi-Points Criterion for Deterministic Parallel Global Optimization Based on Gaussian Processes.”
Wang, Jialei, Clark, C. S, Liu, Eric, Frazier, I. P (2020). “Parallel Bayesian Global Optimization of Expensive Functions.” Operations Research, 68(6), 1850–1865.

Examples

# \donttest{
if (requireNamespace("mlr3learners") &
    requireNamespace("DiceKriging") &
    requireNamespace("rgenoud")) {

  library(bbotk)
  library(paradox)
  library(mlr3learners)

  fun = function(xs) {
    list(y = xs$x ^ 2)
  }
  domain = ps(x = p_dbl(lower = -10, upper = 10))
  codomain = ps(y = p_dbl(tags = "minimize"))
  objective = ObjectiveRFun$new(fun = fun, domain = domain, codomain = codomain)

  instance = OptimInstanceBatchSingleCrit$new(
    objective = objective,
    terminator = trm("evals", n_evals = 7))

  surrogate = default_surrogate(instance)

  acq_function = acqf("ei")

  acq_optimizer = acqo(
    optimizer = opt("random_search", batch_size = 100),
    terminator = trm("evals", n_evals = 100))

  optimizer = opt("mbo",
    loop_function = bayesopt_mpcl,
    surrogate = surrogate,
    acq_function = acq_function,
    acq_optimizer = acq_optimizer,
    args = list(q = 3))

  optimizer$optimize(instance)
}
#>           x  x_domain        y
#>       <num>    <list>    <num>
#> 1: 2.053603 <list[1]> 4.217285
# }