muq

Markov Chain Monte Carlo algorithms from the MUQ package

Topics

bayesian_calibration

Specification

• Alias: None

• Arguments: None

Child Keywords:

Required/Optional	Description of Group	Dakota Keyword	Dakota Keyword Description
Required		chain_samples	Number of Markov Chain Monte Carlo posterior samples
Optional		seed	Seed of the random number generator
Optional		rng	Selection of a random number generator
Optional (Choose One)	MCMC Algorithm	dram	Use the DRAM MCMC algorithm
		delayed_rejection	Use the Delayed Rejection MCMC algorithm
		adaptive_metropolis	Use the Adaptive Metropolis MCMC algorithm
		metropolis_hastings	Use the Metropolis-Hastings MCMC algorithm
		mala	Metropolis-adjusted Langevin algorithm
Optional		dr_num_stages	Number of stages
Optional		dr_scale_type	Type of scaling to use
Optional		dr_scale	Scaling parameter
Optional		am_period_num_steps	Number of steps between updates of the proposal covariance
Optional		am_starting_step	Number of steps prior to start of proposal covariance adaptation
Optional		am_scale	Sample covariance scaling used to define proposal covariance
Optional		mala_step_size	Parameter for the MALA MCMC algorithm in MUQ
Optional		proposal_covariance	Defines the technique used to generate the MCMC proposal covariance.

Description

The muq method supports the following MCMC algorithms: adaptive metropolis (AM), Metropolis Hasting (MH), delayed rejection (DR), delayed-rejection adaptive metropolis (DRAM), or Metropolis-adjusted Langevin algorithm (MALA).

The muq method is currently an experimental method that relies on algorithms from MIT’s MUQ code documented at: https://bitbucket.org/mituq/muq2/src/master/