Enum BradleyTerryOptimizer 

pub enum BradleyTerryOptimizer {
    NewtonRaphson {
        max_iterations: usize,
        tolerance: f64,
        prior_lambda: f64,
    },
    MM {
        max_iterations: usize,
        tolerance: f64,
        bootstrap_samples: usize,
    },
}

Bradley-Terry optimizer configuration

Variants

NewtonRaphson

Newton-Raphson optimization with Fisher Information for uncertainty

Faster convergence, provides analytical covariance matrix from the inverse Fisher Information (negative Hessian).

Fields

max_iterations: usize

Maximum iterations (default: 100)

tolerance: f64

Convergence tolerance for gradient norm (default: 1e-8)

prior_lambda: f64

Gaussian prior precision on the log-strengths (L2 penalty coefficient λ, default: 0.1).

This is a genuine MAP prior: the penalized objective is LL(θ) − (λ/2)·‖θ‖², so the prior contributes −λθ to the gradient and −λ to the Hessian diagonal. It shrinks the log-strengths toward 0 (strength 1), which keeps candidates that win or lose all of their comparisons finite instead of diverging to ±∞.

MM

MM (Minorization-Maximization) algorithm with bootstrap for uncertainty

Simpler, guaranteed monotonic likelihood increase, uses bootstrap resampling to estimate variance.

Fields

max_iterations: usize

Maximum iterations (default: 100)

tolerance: f64

Convergence tolerance for parameter change (default: 1e-8)

bootstrap_samples: usize

Number of bootstrap samples for variance estimation (default: 100)

Implementations

impl BradleyTerryOptimizer

pub fn newton_raphson( max_iterations: usize, tolerance: f64, prior_lambda: f64, ) -> Self

Create Newton-Raphson optimizer with custom parameters

prior_lambda is the precision of the Gaussian prior on the log-strengths (see BradleyTerryOptimizer::NewtonRaphson). A value of 0.0 recovers the unregularized MLE (which can diverge for all-win/all-loss candidates); 0.1 is a sensible default.

pub fn mm( max_iterations: usize, tolerance: f64, bootstrap_samples: usize, ) -> Self

Create MM optimizer with custom parameters

Trait Implementations

impl Clone for BradleyTerryOptimizer

fn clone(&self) -> BradleyTerryOptimizer

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for BradleyTerryOptimizer

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

Formats the value using the given formatter.

impl Default for BradleyTerryOptimizer

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for BradleyTerryOptimizer

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for BradleyTerryOptimizer

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.