Struct CalibrationConfig 

pub struct CalibrationConfig {
    pub max_iterations: usize,
    pub seeds_per_iteration: usize,
    pub patience: usize,
    pub min_improvement: f64,
    pub damping: f64,
    pub rollback: RollbackPolicy,
}

Loop-level config.

Fields

max_iterations: usize

Maximum iterations before the loop gives up. Default 20.

seeds_per_iteration: usize

Seeds evaluated per iteration to amortise the T3 single-shard composite CV ≈ 25 %. Default 3 — matches the v5.31 T3 methodology finding (memory note project_v5_31_overnight_loop).

patience: usize

Convergence patience — stop when the best mean loss hasn’t improved by > min_improvement × σ across patience consecutive iterations. Default 3.

min_improvement: f64

Minimum relative improvement (in units of σ_loss) needed to credit a step as “better”. Default 1.0 — i.e. > 1 standard deviation of the aggregate.

damping: f64

Damping applied to proposed Δ before clipping. Default 0.5 (take half the proposed step to avoid oscillation).

rollback: RollbackPolicy

How to handle a step that worsens the loss.

Trait Implementations

impl Clone for CalibrationConfig

fn clone(&self) -> CalibrationConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for CalibrationConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for CalibrationConfig

fn default() -> Self

Returns the “default value” for a type.