dream
DREAM (DiffeRential Evolution Adaptive Metropolis)
Topics
bayesian_calibration
Specification
Alias: None
Arguments: None
Child Keywords:
Required/Optional |
Description of Group |
Dakota Keyword |
Dakota Keyword Description |
---|---|---|---|
Required |
Number of Markov Chain Monte Carlo posterior samples |
||
Optional |
Seed of the random number generator |
||
Optional |
Number of chains in DREAM |
||
Optional |
Number of candidate points for each crossover. |
||
Optional |
Number of chains used in crossover. |
||
Optional |
Convergence tolerance for the Gelman-Rubin statistic |
||
Optional |
Number of generations a long jump step is taken |
||
Optional |
Use an emulator or surrogate model to evaluate the likelihood function |
||
Optional |
Perform Bayesian inference in standardized probability space |
||
Optional |
Export the MCMC chain to the specified filename |
Description
The DiffeRential Evolution Adaptive Metropolis algorithm is described
in [VtBD+09]. For the DREAM method, one can define the number of
chains used with chains
(minimum 3). The total number of
generations per chain in DREAM is the number of samples ( samples
)
divided by the number of chains ( chains
). The number of chains
randomly selected to be used in the crossover each time a crossover
occurs is crossover_chain_pairs
. There is an extra adaptation
during burn-in, in which DREAM estimates a distribution of crossover
probabilities that favors large jumps over smaller ones in each of the
chains. Normalization is required to ensure that all of the input
dimensions contribute equally. In this process, a discrete number of
candidate points for each crossover value is generated. This parameter
is num_cr
. The gr_threshold
is the convergence tolerance for
the Gelman-Rubin statistic which will govern the convergence of the
multiple chain process. The integer jump_step
forces a long jump
every jump_step
generations. For more details about these
parameters, see [VtBD+09].
Attention: While the emulator
specification for DREAM
includes the keyword posterior_adaptive, it is not yet operational.