Struct Xgcf 

pub struct Xgcf {
    pub node_type: Vec<XgcfNodeType>,
    pub child_offsets: Vec<u32>,
    pub child_indices: Vec<u32>,
    pub lit: Vec<i32>,
    pub decision_var: Vec<u32>,
    pub decision_child_false: Vec<u32>,
    pub decision_child_true: Vec<u32>,
    pub roots: Vec<u32>,
    pub level_offsets: Vec<u32>,
    pub level_nodes: Vec<u32>,
}

Host-side structure-of-arrays representation of a compiled probabilistic circuit.

Fields

node_type: Vec<XgcfNodeType>

Opcode for every node.

child_offsets: Vec<u32>

CSR offsets into Self::child_indices for each node.

child_indices: Vec<u32>

Flattened children of conjunction and disjunction nodes.

lit: Vec<i32>

Signed DIMACS literal for each literal node, or zero otherwise.

decision_var: Vec<u32>

Decision variable for each decision node, or zero otherwise.

decision_child_false: Vec<u32>

False-branch child for each decision node.

decision_child_true: Vec<u32>

True-branch child for each decision node.

roots: Vec<u32>

Circuit root node indices.

level_offsets: Vec<u32>

CSR offsets into Self::level_nodes for bottom-up evaluation levels.

level_nodes: Vec<u32>

Node indices ordered by bottom-up evaluation level.

Implementations

impl Xgcf

pub fn from_ddnnf(ddnnf: &DecisionDnnf) -> Result<Self>

Converts a decision-DNNF circuit into XGCF and validates its topology.

pub fn smooth_random_vars(&self, is_random_var: &[bool]) -> Result<Self>

Return a semantically equivalent circuit that is smooth with respect to the subset of variables marked as random in is_random_var.

Smoothness guarantees that, for any OR/DECISION node, all branches mention the same set of random variables. This makes WMC evaluation and gradients correct even when evidence/queries force random variables in non-smooth Decision-DNNF outputs.

pub fn eval_log_wmc<F>(&self, var_log_weights: F) -> Result<f64>

where F: Fn(u32) -> (f64, f64),

Evaluates the circuit’s weighted model count in log space on the CPU.

pub fn eval_log_wmc_and_grads( &self, var_log_weights: &[(f64, f64)], ) -> Result<(f64, Vec<f64>, Vec<f64>)>

Evaluates log weighted model count and literal log-weight gradients on the CPU.

Trait Implementations

impl Clone for Xgcf

fn clone(&self) -> Xgcf

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Xgcf

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

Formats the value using the given formatter.