wasabi

(Experimental Method) Non-MCMC Bayesian inference using interval analysis

Specification

  • Alias: None

  • Arguments: None

  • JSON argument key: None

Child Keywords:

Required/Optional

Description of Group

JSON Group Key

Dakota Keyword

Dakota Keyword Description

Optional

emulator

Use an emulator or surrogate model to evaluate the likelihood function

Required

pushforward_samples

(Experimental Capability) Number of samples of the prior to push forward through the model

Optional

seed

Seed of the random number generator

Optional

standardized_space

Perform Bayesian inference in standardized probability space

Required

data_distribution

(Experimental Capability) Specify the distribution of the experimental data

Optional

posterior_samples_import_filename

(Experimental Capability) Filename for samples at which the user would like the posterior density calculated

Optional

generate_posterior_samples

(Experimental Capability) Generate random samples from the posterior density

Optional

evaluate_posterior_density

(Experimental Capability) Evaluate the posterior density and output to the specified file

Description

Offers an alternative to Markov Chain Monte Carlo-based Bayesian inference. This is a nascent capability, not yet ready for production use.

Usage Guidelines: The WASABI method requires an emulator model.

Attention: While the emulator specification for WASABI includes the keyword posterior_adaptive, it is not yet operational.

Examples

method
  bayes_calibration
    wasabi
      pushforward_samples = 100
      data_distribution gaussian
        means = 0.0
        covariance diagonal values 1.0
{
  "method": [
    {
      "bayes_calibration": {
        "sub_method": {
          "wasabi": {
            "pushforward_samples": 100,
            "data_distribution": {
              "gaussian": {
                "means": [0.0],
                "covariance": {
                  "values": [1.0],
                  "type": {
                    "diagonal": true
                  }
                }
              }
            }
          }
        }
      }
    }
  ]
}