Dakota Reference Manual
Version 6.4
LargeScale Engineering Optimization and Uncertainty Analysis

Use adaptive basis selection to choose the basis terms in a polynomial chaos expansion.
Alias: none
Argument(s): none
Required/Optional  Description of Group  Dakota Keyword  Dakota Keyword Description  

Optional  advancements  The maximum number of steps used to expand a basis step.  
Optional  soft_convergence_limit  The maximum number of times no improvement in cross validation error is allowed before the algorithm is terminated. 
Use adaptive basis selection to choose the basis terms in a polynomial chaos expansion. Basis selection uses compressed sensing to identify a initial set of non zero PCE coefficients. Then these nonzero terms are expanded a set number of times (we suggest 3) and compressed sensing is then applied to these three new index sets. Cross valiation is used to choose the best candidate basis. The best basis is then restricted again to the nonzero terms and expanded until no improvement can be gained by adding additional terms.