Struct RobustPositiveEvidencePoolOperator 

pub struct RobustPositiveEvidencePoolOperator {
    pub signals: Vec<Arc<dyn Operator>>,
    pub alpha: f64,
    pub gating: LogitGating,
    pub base_rate: Option<f64>,
    pub weights: Option<Vec<f64>>,
    pub adaptive_weights: bool,
    pub logit_min: Option<Vec<f64>>,
    pub logit_max: Option<Vec<f64>>,
    pub top_k: Option<usize>,
}

Robust positive-evidence pooling for retrieval ranking.

Each signal must produce prior-free evidence probabilities in [0, 1]; a configured base_rate enters the pool exactly once. This is a ranking heuristic, not the conditional-independence Bayesian sum implemented by BayesianEvidenceFusionOperator. Missing documents contribute zero gated logit, and a signal with no matches at all stays in the declared signal set as neutral evidence (Lucene PR 16410 semantics: the clause count that governs n^alpha and the uniform denominator never shrinks at execution time). The default softplus gating floors match evidence at the prior; LogitGating::Pass matches Lucene’s signed default.

Fields

signals: Vec<Arc<dyn Operator>>

alpha: f64

gating: LogitGating

base_rate: Option<f64>

weights: Option<Vec<f64>>

adaptive_weights: bool

Derive per-signal weights from each signal’s gated-evidence spread over its matches (Theorem 8.3 reliability weighting, estimated unsupervised). Ignored when explicit weights are set.

logit_min: Option<Vec<f64>>

logit_max: Option<Vec<f64>>

top_k: Option<usize>

Implementations

impl RobustPositiveEvidencePoolOperator

pub fn new(signals: Vec<Arc<dyn Operator>>, alpha: f64) -> Self

pub fn with_adaptive_weights(self) -> Self

pub fn with_gating(self, gating: LogitGating) -> Self

pub fn with_base_rate(self, base_rate: f64) -> Self

Fusion-level relevance prior, applied exactly once.

pub fn with_weights(self, weights: Vec<f64>) -> Self

pub fn with_logit_normalization( self, logit_min: Vec<f64>, logit_max: Vec<f64>, ) -> Self

pub fn with_top_k(self, top_k: usize) -> Self

Trait Implementations

impl Operator for RobustPositiveEvidencePoolOperator

fn execute(&self, ctx: &ExecutionContext) -> OperatorResult

fn cost_estimate(&self, stats: &IndexStats) -> f64