Struct IsotonicRegression 

pub struct IsotonicRegression {
    pub thresholds: Vec<f32>,
    pub values: Vec<f32>,
}

Isotonic Regression calibrator.

Non-parametric calibration method that fits a monotonically increasing step function to the data. Guaranteed to be monotonic.

Reference

• Zadrozny, B., & Elkan, C. (2002). Transforming classifier scores into accurate multiclass probability estimates.

Fields

thresholds: Vec<f32>

Threshold values (sorted)

values: Vec<f32>

Calibrated values at each threshold

Implementations

impl IsotonicRegression

pub fn new() -> IsotonicRegression

Create a new isotonic regression calibrator.

pub fn fit(&mut self, predictions: &[f32], labels: &[bool])

Fit isotonic regression on predictions and labels.

Uses Pool Adjacent Violators (PAV) algorithm.

pub fn predict(&self, prediction: f32) -> f32

Get calibrated probability for a prediction.

pub fn predict_batch(&self, predictions: &[f32]) -> Vec<f32>

Get calibrated probabilities for multiple predictions.

Trait Implementations

impl Clone for IsotonicRegression

fn clone(&self) -> IsotonicRegression

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for IsotonicRegression

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

Formats the value using the given formatter.

impl Default for IsotonicRegression

fn default() -> IsotonicRegression

Returns the “default value” for a type.