Dakota Reference Manual  Version 6.4
Large-Scale Engineering Optimization and Uncertainty Analysis
 All Pages
proposal_updates


Restarts the MCMC chain with updated derivative-based proposal covariance.

Specification

Alias: none

Argument(s): INTEGER

Description

When employing derivative-based proposal covariance, this specification defines the number of restarts that are performed during the course of the total sample size of the MCMC chain. For each restart, a new chain is initiated from the final point in the previous acceptance chain using updated proposal covariance corresponding to the derivatives values at the new starting point.

Default Behavior

If proposal_updates is not specified, then the default frequency for restarting the chain with updated proposal covariance is every 100 samples.

Expected Output

Each restarted chain will generate a new QUESO header and sampling summary, and the chain diagnostics will be appended within the outputData directory.

Usage Tips

proposal_updates should be tailored to the size of the total chain, accounting for the relative expense of derivative-based proposal updates.

Examples

method,
        bayes_calibration queso
          samples = 2000 seed = 348
          delayed_rejection
          emulator pce sparse_grid_level = 2
          proposal_covariance derivatives
            proposal_updates = 50 # restarted chains, each with 40 new points