Enum MechanismSlot 

pub enum MechanismSlot {
    Vacant,
    Pending {
        family_id: Arc<str>,
    },
    LinearGaussian {
        intercept: f64,
        coeffs: Arc<[f64]>,
        sigma: f64,
    },
    Discrete {
        support: Arc<[f64]>,
        probs: Arc<[f64]>,
        logit_coeffs: Option<Arc<[f64]>>,
    },
    Constant {
        value: f64,
    },
    HierarchicalLinear {
        intercept: f64,
        coeffs: Arc<[f64]>,
        sigma: f64,
        shrinkage: f64,
    },
    Bvar {
        intercept: f64,
        coeffs: Arc<[f64]>,
        sigma: f64,
    },
    LinearGaussianStateSpace {
        a: f64,
        process_std: f64,
        obs_std: f64,
        initial_mean: f64,
    },
    GaussianProcess {
        length_scale: f64,
        variance: f64,
        noise_std: f64,
        x_train: Arc<[f64]>,
        n_train: usize,
        n_parents: usize,
        alpha: Arc<[f64]>,
    },
    Dynamic {
        id: Arc<str>,
        mechanism: Arc<dyn DynamicMechanism>,
    },
}

Mechanism slot filled by fitting / registry.

Variants

Vacant

Unassigned.

Pending

Assigned family id pending fit.

Fields

family_id: Arc<str>

Family registry id.

LinearGaussian

Fitted linear Gaussian: intercept + parent coeffs + residual σ.

Fields

intercept: f64

Intercept.

coeffs: Arc<[f64]>

Coefficients aligned with ParentGatherPlan::parents.

sigma: f64

Residual standard deviation.

Discrete

Discrete categorical over finite support.

Unconditional when logit_coeffs is None (use probs). Parent-conditional when logit_coeffs is Some: softmax over support.len() rows of length 1 + n_parents (intercept + parent coeffs). Coefficients are baseline-category multinomial-logit MLEs (reference category index 0 is pinned to zero).

Fields

support: Arc<[f64]>

Support values.

probs: Arc<[f64]>

Unconditional probabilities (same length as support); ignored when logit_coeffs is set.

logit_coeffs: Option<Arc<[f64]>>

Optional softmax logit coefficients, row-major [k * (1 + p)].

Constant

Constant mechanism.

Fields

value: f64

Fixed value.

HierarchicalLinear

Hierarchical linear Gaussian (partial-pooling / ridge toward prior mean 0).

Fields

intercept: f64

Intercept.

coeffs: Arc<[f64]>

Parent coefficients (shrunk).

sigma: f64

Residual standard deviation.

shrinkage: f64

Shrinkage strength used at fit (λ on diagonal of XtX).

Bvar

Bayesian VAR-style linear Gaussian on parent lags (single-equation).

Fields

intercept: f64

Intercept.

coeffs: Arc<[f64]>

Parent / lag coefficients.

sigma: f64

Residual standard deviation.

LinearGaussianStateSpace

Linear Gaussian state-space observation mechanism (1-D LGSSM).

Latent: x_t = a x_{t-1} + σ_proc ε; observation: y_t = x_t + σ_obs η. Parents unused at evaluate time (state evolves from shared noise).

Fields

a: f64

AR coefficient.

process_std: f64

Process noise std.

obs_std: f64

Observation noise std.

initial_mean: f64

Initial latent mean.

GaussianProcess

Gaussian-process mechanism (RBF dual form); requires gaussian-process feature to fit.

Fields

length_scale: f64

Length scale.

variance: f64

Signal variance.

noise_std: f64

Observation noise std.

x_train: Arc<[f64]>

Training parent rows, row-major [n_train * n_parents].

n_train: usize

Training rows.

n_parents: usize

Parent arity.

alpha: Arc<[f64]>

Dual coefficients α = (K + σ²I)^{-1} y.

Dynamic

Explicit slow-path dynamic / user mechanism (not serializable).

Fields

id: Arc<str>

Stable label for diagnostics (e.g. variable name).

mechanism: Arc<dyn DynamicMechanism>

Object-safe mechanism implementation.

Trait Implementations

impl Clone for MechanismSlot

fn clone(&self) -> MechanismSlot

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for MechanismSlot

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

Formats the value using the given formatter.

impl Default for MechanismSlot

fn default() -> MechanismSlot

Returns the “default value” for a type.