trustRegion: Performs trust region refinement

In SACOBRA: Self-Adjusting COBRA

Description

If cobra$TrustRegion==TRUE (see cobraInit), then the trustRegion functionality is applied every iteration in order to refine the best solution so far. This function builds a local model around the best solution and runs a local search in the trust region to refine the best solution and find a better solution in the neighborhood.

Usage

trustRegion(cobra, center = cobra$xbest)

Arguments

cobra	an object of class cobra, which is basically a list (see cobraInit)
center	[cobra$xbest] the center of the trust region

Value

the modified cobra with new/updated elements

TRDONE	logical, is TRUE if there are more than d+1 points in the trusted region and thus surrogates can be trained. Otherwise FALSE.
trustregX	if TRDONE==TRUE the refined solution from the trust-region call, otherwise NA

If TRDONE==TRUE the relevant lists and counters (A,Fres,df,...) of cobra will be updated in cobraPhaseII as well.

Author(s)

Samineh Bagheri ([e-mail])

SACOBRA documentation built on March 26, 2020, 7:15 p.m.