Struct MultiConfusionMatrix

pub struct MultiConfusionMatrix {
    pub dim: usize,
    pub counts: Vec<Vec<usize>>,
    /* private fields */
}

Confusion matrix for multi-class classification, in which rows represent predicted counts and columns represent labeled counts

Fields

dim: usize

output dimension

counts: Vec<Vec<usize>>

count data

Implementations

impl MultiConfusionMatrix

pub fn compute<T: Scalar>( scores: &Vec<Vec<T>>, labels: &Vec<usize>, ) -> Result<MultiConfusionMatrix, EvalError>

Computes a new confusion matrix from the provided scores and labels

Arguments

• scores - vector of class scores
• labels - vector of class labels (indexed at zero)

Errors

An invalid input error will be returned if either scores or labels are empty, or if their lengths do not match. An undefined metric error will be returned if scores contain any value that is not finite.

Examples

use eval_metrics::classification::MultiConfusionMatrix;
let scores = vec![
   vec![0.3, 0.1, 0.6],
   vec![0.5, 0.2, 0.3],
   vec![0.2, 0.7, 0.1],
   vec![0.3, 0.3, 0.4],
   vec![0.5, 0.1, 0.4],
   vec![0.8, 0.1, 0.1],
   vec![0.3, 0.5, 0.2]
];
let labels = vec![2, 1, 1, 2, 0, 2, 0];
let matrix = MultiConfusionMatrix::compute(&scores, &labels)?;

pub fn from_counts( counts: Vec<Vec<usize>>, ) -> Result<MultiConfusionMatrix, EvalError>

Constructs a multi confusion matrix with the provided counts

Arguments

• counts - vector of vector of counts, where each inner vector represents a row in the confusion matrix

Errors

An invalid input error will be returned if the counts are not a square matrix, or if the counts are all zero

Examples

use eval_metrics::classification::MultiConfusionMatrix;
let counts = vec![
    vec![8, 3, 2],
    vec![1, 5, 3],
    vec![2, 1, 9]
];
let matrix = MultiConfusionMatrix::from_counts(counts)?;

pub fn accuracy(&self) -> Result<f64, EvalError>

Computes accuracy

pub fn precision(&self, avg: &Averaging) -> Result<f64, EvalError>

Computes precision, which necessarily requires a specified averaging method

Arguments

• avg - averaging method, which can be either ‘Macro’ or ‘Weighted’

pub fn recall(&self, avg: &Averaging) -> Result<f64, EvalError>

Computes recall, which necessarily requires a specified averaging method

Arguments

• avg - averaging method, which can be either ‘Macro’ or ‘Weighted’

pub fn f1(&self, avg: &Averaging) -> Result<f64, EvalError>

Computes F1, which necessarily requires a specified averaging method

Arguments

• avg - averaging method, which can be either ‘Macro’ or ‘Weighted’

pub fn rk(&self) -> Result<f64, EvalError>

Computes Rk, also known as the multi-class Matthews correlation coefficient following the approach of Gorodkin in “Comparing two K-category assignments by a K-category correlation coefficient” (2004)

pub fn per_class_accuracy(&self) -> Vec<Result<f64, EvalError>>

Computes per-class accuracy, resulting in a vector of values for each class

pub fn per_class_precision(&self) -> Vec<Result<f64, EvalError>>

Computes per-class precision, resulting in a vector of values for each class

pub fn per_class_recall(&self) -> Vec<Result<f64, EvalError>>

Computes per-class recall, resulting in a vector of values for each class

pub fn per_class_f1(&self) -> Vec<Result<f64, EvalError>>

Computes per-class F1, resulting in a vector of values for each class

pub fn per_class_mcc(&self) -> Vec<Result<f64, EvalError>>

Computes per-class MCC, resulting in a vector of values for each class

Trait Implementations

impl Clone for MultiConfusionMatrix

fn clone(&self) -> MultiConfusionMatrix

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for MultiConfusionMatrix

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

Formats the value using the given formatter.

impl Display for MultiConfusionMatrix

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

Formats the value using the given formatter.

impl PartialEq for MultiConfusionMatrix

fn eq(&self, other: &MultiConfusionMatrix) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Eq for MultiConfusionMatrix

impl StructuralPartialEq for MultiConfusionMatrix