Struct DisturbanceClassifier 

pub struct DisturbanceClassifier {
    pub excess_threshold: f32,
    pub persistence_min: u32,
    pub drift_lambda_min: f32,
    /* private fields */
}

Heuristic disturbance classifier.

Given observable quantities from the grammar/DSA/Lyapunov pipeline, produces a candidate RfDisturbance hypothesis with a confidence score.

This is a structural classifier — it operates on the shape of the residual trajectory, not on modulation features. It is therefore modulation-agnostic by construction.

Decision rules

The rules are derived from the DDMF disturbance model signatures:

Observation	Likely disturbance
Large λ + sustained outward drift	Drift
Abrupt step in ‖r‖ with sustained elevation	PersistentElevated
Single spike above ρ then return	Impulsive
Slowly increasing ‖r̈‖ trend	SlewRateBounded
Stationary bounded noise	PointwiseBounded

Fields

excess_threshold: f32

Threshold: normalised-excess (‖r‖−ρ)/ρ above which a sample is “notably outside.”

persistence_min: u32

Minimum consecutive samples above threshold to classify as PersistentElevated.

drift_lambda_min: f32

Lyapunov λ threshold below which Drift is not inferred.

Implementations

impl DisturbanceClassifier

pub const fn default_rf() -> Self

Construct with default RF thresholds.

pub fn classify( &mut self, norm: f32, rho: f32, lambda: f32, dsa_fired: bool, ) -> Option<DisturbanceHypothesis>

Classify one observation.

• norm: current ‖r(k)‖
• rho: admissibility envelope radius
• lambda: Lyapunov exponent from lyapunov module (pass 0.0 if unknown)
• dsa_fired: whether the DSA motif-fired flag is active

Returns Some(DisturbanceHypothesis) when a classification is made; None during nominal operation.

pub fn reset(&mut self)

Reset internal state.