KSLCore/ksl.simopt.solvers.algorithms.bo/SampledAcquisitionOptimizer

SampledAcquisitionOptimizer

class SampledAcquisitionOptimizer(numCandidates: Int = BayesianOptimizationSolver.defaultNumCandidates, numLocalRestarts: Int = BayesianOptimizationSolver.defaultRestarts, localFraction: Double = 0.1) : AcquisitionOptimizerIfc(source)

A robust, dependency-free acquisition optimizer: it scores a Latin-hypercube candidate set over the feasible region plus a set of local candidates around the current best, and returns the best-scoring candidate (rounded to granularity via the problem definition). Because acquisition evaluation is cheap (no simulation), a large candidate set gives good coverage.

Parameters

numCandidates

the number of global Latin-hypercube candidates. Must be >= 1. Defaults to BayesianOptimizationSolver.defaultNumCandidates.

numLocalRestarts

the number of local candidates sampled around the current best. Must be

= 0. Defaults to BayesianOptimizationSolver.defaultRestarts.

localFraction

the standard deviation of local perturbations, as a fraction of each input range. Must be > 0. Defaults to 0.1.

Constructors

SampledAcquisitionOptimizer

constructor(numCandidates: Int = BayesianOptimizationSolver.defaultNumCandidates, numLocalRestarts: Int = BayesianOptimizationSolver.defaultRestarts, localFraction: Double = 0.1)

Properties

localFraction

var localFraction: Double

numCandidates

var numCandidates: Int

numLocalRestarts

var numLocalRestarts: Int

Functions

optimize

open override fun optimize(acq: AcquisitionFunctionIfc, surrogate: SurrogateModelIfc, incumbent: Double, bo: BayesianOptimizationSolver): InputMap

Returns the input point that (approximately) maximizes acq over the feasible region.

toString

open override fun toString(): String