![]() |
Dakota Reference Manual
Version 6.15
Explore and Predict with Confidence
|
Order for tensor-products of Gaussian quadrature rules
Alias: none
Argument(s): INTEGER
Child Keywords:
Required/Optional | Description of Group | Dakota Keyword | Dakota Keyword Description | |
---|---|---|---|---|
Optional | dimension_preference | A set of weights specifying the realtive importance of each uncertain variable (dimension) | ||
Optional (Choose One) | Quadrature Rule Nesting (Group 1) | nested | Enforce use of nested quadrature rules if available | |
non_nested | Enforce use of non-nested quadrature rules |
Multidimensional integration by a tensor-product of Gaussian quadrature rules (specified with quadrature_order
, and, optionally, dimension_preference
). The default rule selection is to employ non_nested
Gauss rules including Gauss-Hermite (for normals or transformed normals), Gauss-Legendre (for uniforms or transformed uniforms), Gauss-Jacobi (for betas), Gauss-Laguerre (for exponentials), generalized Gauss-Laguerre (for gammas), and numerically-generated Gauss rules (for other distributions when using an Extended basis). For the case of p_refinement
or the case of an explicit nested
override, Gauss-Hermite rules are replaced with Genz-Keister nested rules and Gauss-Legendre rules are replaced with Gauss-Patterson nested rules, both of which exchange lower integrand precision for greater point reuse. By specifying a dimension_preference
, where higher preference leads to higher order polynomial resolution, the tensor grid may be rendered anisotropic. The dimension specified to have highest preference will be set to the specified quadrature_order
and all other dimensions will be reduced in proportion to their reduced preference; any non-integral portion is truncated. To synchronize with tensor-product integration, a tensor-product expansion is used, where the order of the expansion in each dimension is selected to be half of the integrand precision available from the rule in use, rounded down. In the case of non-nested Gauss rules with integrand precision
,
is one less than the quadrature order
in each dimension (a one-dimensional expansion contains the same number of terms,
, as the number of Gauss points). The total number of terms, N, in a tensor-product expansion involving n uncertain input variables is
In some advanced use cases (e.g., multifidelity UQ), multiple grid resolutions can be employed; for this reason, the quadrature_order
specification supports an array input.
A corresponding sequence specification is documented at, e.g., quadrature_order_sequence and quadrature_order_sequence
These keywords may also be of interest: