Struct BayesianNetworkInference 

pub struct BayesianNetworkInference { /* private fields */ }

Bayesian network inference engine.

Supports variable elimination (exact), belief propagation (sum-product), and likelihood-weighted sampling.

Implementations

impl BayesianNetworkInference

pub fn new( network: BayesianNetwork, config: BniConfig, ) -> Result<Self, BniError>

Create and validate a new inference engine.

Returns an error if:

• the network contains a directed cycle,
• any CPT references an undefined variable,
• any CPT has incorrect dimensions.

pub fn query( &mut self, q: &InferenceQuery, ) -> Result<Vec<QueryResult>, BniError>

Run an inference query and return per-variable posterior distributions.

pub fn prior_marginal( &mut self, variable: &str, ) -> Result<QueryResult, BniError>

Compute the prior marginal for a single variable (no evidence).

pub fn add_cpt( &mut self, cpt: ConditionalProbabilityTable, ) -> Result<(), BniError>

Add a new CPT to the network after construction.

Validates dimensions and checks for duplicate variables.

pub fn d_separated( &self, x: &str, y: &str, z: &[String], ) -> Result<bool, BniError>

Test d-separation: are x and y d-separated given observations z (the Bayes Ball algorithm)?

Returns true if x ⊥ y | z.

pub fn stats(&self) -> BniStats

Return a snapshot of engine statistics.