One trust-region round's outcome

emulatorTrustStep(state, improvement, tol, ctl, atEdge = FALSE)

Arguments

state

list: scale, successes, failures.

improvement

numeric; the GP's best mean before the round minus after it.

tol

numeric; an improvement above it is a success.

ctl

emulatorControlOf().

atEdge

logical; the GP's best point after the round is in the round's box margin (emulatorAtBoxEdge()): a success then does not shrink the box.

Value

state updated, with stop (logical) and reason.