Skip to main content

box_robust

scirs2_optimize::robust

Function box_robust 

Source 
pub fn box_robust<F>(
    f: &F,
    x: &ArrayView1<'_, f64>,
    delta: &ArrayView1<'_, f64>,
) -> OptimizeResult<f64>
where
    F: Fn(&ArrayView1<'_, f64>) -> f64,
Expand description

Evaluate the worst-case objective over an axis-aligned box uncertainty set.

For each dimension i, the perturbed parameter x̃_i is chosen from {x_i - δ_i, x_i + δ_i} to maximise f. The full worst-case is approximated by multi-start local search over the 2n vertex candidates plus random interior samples.

§Arguments

  • f – objective function (lower is better; we find the maximum)
  • x – nominal parameter vector (length n)
  • delta – per-component box radius (length n)

§Returns

The worst-case value max_{ξ : |ξ_i| ≤ δ_i} f(x + ξ).