posterior_adaptive

Adapt emulator model to increase accuracy in high posterior probability regions

Specification

  • Alias: None

  • Arguments: None

Description

Following an emulator-based MCMC process, this option refines the emulator by selecting points in regions of high posterior probability, performing truth evaluations at these points, updating the emulator, and reperforming the MCMC process. The adaptation is continued until the maximum number of iterations is exceeded or the convergence tolerance is met.

Examples

bayes_calibration queso
  chain_samples = 2000 seed = 348
  delayed_rejection
  emulator
    gaussian_process surfpack build_samples = 30
    posterior_adaptive max_iterations = 10
    proposal_covariance derivatives