Struct CmaEsTolerance 

pub struct CmaEsTolerance<F = f64> { /* private fields */ }

CMA-ES TolX convergence test (Hansen 2016 Appendix B.3): stop when the largest standard deviation of any axis of the search distribution drops below tol_x, i.e. σ · maxᵢ dᵢ < tol_x (where dᵢ are the square roots of C’s eigenvalues).

Binds on the concrete CmaEsState — the canonical convergence criterion for both CmaEs and BoundedCmaEs, which share that state. The Hansen-recommended default is 1e−12 · initial_sigma (scaled by maxᵢ stdsᵢ when an anisotropic initial covariance is used, to stay relative to the initial spread).

Implementations

impl<F> CmaEsTolerance<F>

pub fn new(tol_x: F) -> Self

New TolX criterion firing at σ · maxᵢ dᵢ < tol_x.

Trait Implementations

impl<V, M, F> TerminationCriterion<CmaEsState<V, M, F>> for CmaEsTolerance<F>

where F: Scalar, V: VectorLen + Index<usize, Output = F>,

fn check(&mut self, state: &CmaEsState<V, M, F>) -> Option<TerminationReason>

Inspect the current state and return Some(reason) to halt the run, or None to continue. Called once per iteration before the solver’s next_iter.

fn reset(&mut self)

Reset any internal per-run state so the criterion behaves as if freshly constructed.