Struct DynamicBayesianNetwork 

pub struct DynamicBayesianNetwork {
    pub state_vars: Vec<(String, usize)>,
    pub observation_vars: Vec<(String, usize)>,
    pub initial_dists: HashMap<String, ArrayD<f64>>,
    pub transition_dists: HashMap<String, ArrayD<f64>>,
    pub emission_dists: HashMap<String, ArrayD<f64>>,
}

Dynamic Bayesian Network.

Represents a temporal probabilistic model with variables that evolve over time.

Example

use tensorlogic_quantrs_hooks::DynamicBayesianNetwork;
use scirs2_core::ndarray::{Array, ArrayD, IxDyn};

// Create a simple DBN with one state variable
let dbn = DynamicBayesianNetwork::new(
    vec![("state".to_string(), 2)],  // state variables with cardinality
    vec![],  // no observation variables
);

Fields

state_vars: Vec<(String, usize)>

State variables with cardinalities

observation_vars: Vec<(String, usize)>

Observation variables with cardinalities

initial_dists: HashMap<String, ArrayD<f64>>

Initial distribution P(X_0) for each state variable

transition_dists: HashMap<String, ArrayD<f64>>

Transition distributions P(X_t | X_{t-1})

emission_dists: HashMap<String, ArrayD<f64>>

Emission distributions P(Y_t | X_t)

Implementations

impl DynamicBayesianNetwork

pub fn new( state_vars: Vec<(String, usize)>, observation_vars: Vec<(String, usize)>, ) -> Self

Create a new DBN with state and observation variables.

pub fn set_initial(&mut self, var: &str, dist: ArrayD<f64>) -> Result<&mut Self>

Set initial distribution for a state variable.

pub fn set_transition( &mut self, var: &str, dist: ArrayD<f64>, ) -> Result<&mut Self>

Set transition distribution P(X_t | X_{t-1}) for a state variable.

pub fn set_emission( &mut self, obs_var: &str, dist: ArrayD<f64>, ) -> Result<&mut Self>

Set emission distribution P(Y | X) for an observation variable.

pub fn unroll(&self, num_steps: usize) -> Result<FactorGraph>

Unroll the DBN for a fixed number of time steps.

Returns a FactorGraph representing the unrolled DBN.

pub fn filter( &self, observations: &[HashMap<String, usize>], ) -> Result<Vec<HashMap<String, ArrayD<f64>>>>

Perform filtering to compute P(X_t | y_{1:t}).

Returns marginal distributions for state variables at each time step.

pub fn smooth( &self, observations: &[HashMap<String, usize>], ) -> Result<Vec<HashMap<String, ArrayD<f64>>>>

Perform smoothing to compute P(X_t | y_{1:T}) for all t.

Uses variable elimination on the unrolled DBN.

pub fn viterbi( &self, observations: &[HashMap<String, usize>], ) -> Result<Vec<HashMap<String, usize>>>

Compute most likely sequence using Viterbi algorithm on unrolled DBN.

pub fn state_cardinalities(&self) -> HashMap<String, usize>

Get interface (state) variable cardinalities.

pub fn observation_cardinalities(&self) -> HashMap<String, usize>

Get observation variable cardinalities.

pub fn all_variables(&self) -> HashSet<String>

Get all variables in the DBN.

pub fn run_belief_propagation( &self, num_steps: usize, evidence: &HashMap<String, usize>, ) -> Result<HashMap<String, ArrayD<f64>>>

Run belief propagation on the unrolled DBN.

Trait Implementations

impl Clone for DynamicBayesianNetwork

fn clone(&self) -> DynamicBayesianNetwork

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DynamicBayesianNetwork

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

Formats the value using the given formatter.