Emulators API
The EmulationBasedInference module defines types and methods for emulating expensive SimulatorForwardProblems.
EmulationBasedInference.Decorrelated — Type
Decorrelated{bijType} <: DataTransformDecorrelation transform applied to a multivariate output (target) space.
EmulationBasedInference.EmulatedObservables — Type
EmulatedObservables{names} <: SciMLBase.AbstractSciMLAlgorithmRepresents an emulated forward solver for observables names in a given forward problem. The dimensionality of the output spaces of the emulators are assumed to match those of the observables exactly.
EmulationBasedInference.Emulator — Type
Emulator{TM,TT} <: EmulatorData structure consisting of some training data for a model emulator, appropriate transforms, and a tuple of univariate regressors which are applied to the transformed data.
EmulationBasedInference.EmulatorData — Type
EmulatorDataGeneric container for emulator training data matrices X and Y. X should have shape m x N where N is the number of samples and m is the number of covariates. Y should have dimensions N x d where d is the number of ouptut covariates.+
EmulationBasedInference.GPRegressor — Type
GPRegressorGeneric implementation of a Gaussian Process regressor for a univariate outputs. This implementation is adapted from MLJGaussianProcesses to be usable without MLJ.
EmulationBasedInference.StackedRegressors — Type
StackedRegressors{TM}Simple representation of a set of stacked, univaraite regressors for multi-target regression problems.