API

BayesModel(prior, logintegrand) <: AbstractBQModel

Model inheriting from AbstractMCMC.AbstractModel. prior should be a multivariate distribution from Distributions.jl at the moment prior has to be a MvNormal but this will improved in a later version logintegrand should be the log of the function to integrate.

PriorSampling()

Sampler which will use the prior distribution from the given model to provide samples.

BayesQuad(k::Kernel; l=1.0, σ::Real=1.0)

Tool for estimating the bayesian quadrature for a given BayesQuadModel and a Sampler.

BayesQuad estimate the probability distribution p(I) of I = ∫ f(x) p(x) dx

Argument

• k::Kernel : kernel from KernelFunctions.jl, the kernel cannot be composite

Keywords argument

• l : lengthscale of the kernel. It can be:
  • Real : isotropic kernel
  • AbstractVector : ARD kernel (one lengthscale per dimension)
  • LowerTriangular : Linear transformation of the inputs
• σ : variance of the kernel (k(x,x') -> σ * k(x,x'))

Note: If k already has a variance and/or a transformation, these will be automatically extracted and replace the given keyword arguments