Development build. This page describes main, not a released package. The latest published Lodestar.Metrics is 0.3.0 — read its documentation.

Home › Metrics › Classification metrics

The fraction of samples predicted wrongly — sklearn.metrics.zero_one_loss.

public static double Score(ReadOnlySpan<int> yTrue, ReadOnlySpan<int> yPred, bool normalize = true, ReadOnlySpan<double> sampleWeight = default)
public static double Score(ReadOnlySpan<bool> yTrue, ReadOnlySpan<bool> yPred, int labelCount, bool normalize = true, ReadOnlySpan<double> sampleWeight = default)

Parameters — the first overload takes yTrue and yPred as one label per sample, the second as a row-major label matrix with labelCount values per row. normalize divides by the total weight; false returns the weight of the wrong samples instead. sampleWeight is one weight per sample, or empty.

Returns — double in [0, 1] when normalize holds, and the weight of the wrong samples when it does not — a count when every weight is 1.

Exceptions — ArgumentException when the inputs disagree in length, are empty, when the matrix is not a whole number of rows of labelCount, or when the weights do not match the sample count. A sampleWeight holding NaN or an infinity is refused with "Input sample_weight contains NaN." or its infinity counterpart, and one that is zero throughout with "Sample weights must contain at least one non-zero number." — both ArgumentException naming sampleWeight, as scikit-learn's _check_sample_weight refuses them. Weights that merely sum to zero are refused too, with numpy's "Weights sum to zero, can't be normalized." — only while normalize is true, since the weight of the wrong samples never divides.

Example — the count, which is what normalize: false is for.

using Lodestar.Metrics;

int[] truth = [0, 1, 2, 1];
int[] predicted = [0, 2, 2, 1];

double wrong = ZeroOneLoss.Score(truth, predicted, normalize: false);  // => 1

Remarks — with weights and normalize: false the answer is the weight of the wrong samples, not how many there are: 2 on the example above under weights [1, 2, 3, 4], where the count is 1. That is the reference's behaviour and the same widening TopKAccuracy.Score documents for its own normalize.

Applies to — net10.0, netstandard2.0.

See also — HammingLoss.Score, Accuracy.Score, the Python equivalence table.