Metrics pinballloss peroutput

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

PinballLoss.PerOutput

One pinball loss per output, unreduced.

public static double[] PerOutput(ReadOnlySpan<double> yTrue, ReadOnlySpan<double> yPred, double alpha = 0.5, int outputCount = 1, ReadOnlySpan<double> sampleWeight = default)

Parameters — yTrue and yPred are the true and predicted values, row-major. alpha is the quantile being scored, in [0, 1]. outputCount is how many outputs each row holds, and sampleWeight weights the rows.

Returns — a fresh double[] of outputCount entries, in column order.

Exceptions — ArgumentException when a length disagrees with the shape, the input is empty, or it holds a non-finite value; ArgumentOutOfRangeException when outputCount is below one, or alpha is outside [0, 1] — including NaN.

Example — three samples, two outputs, at the median.

using Lodestar.Metrics;

double[] yTrue = [0.5, 1.0, -1.0, 1.0, 7.0, -6.0];
double[] yPred = [0.0, 2.0, -1.0, 2.0, 8.0, -5.0];

double[] perOutput = PinballLoss.PerOutput(yTrue, yPred, alpha: 0.5, outputCount: 2);
double first = perOutput[0];    // => 0.25
double second = perOutput[1];   // => 0.5

Remarks — the array is what you want when a model emits several quantiles of the same target and you laid them out as outputs. It is also the one place the shared multioutput shape is slightly awkward: alpha is a single value applied to every output, so scoring a 10th, a 50th and a 90th percentile means three calls rather than one.

The trap is exactly that. A single call with three quantile columns and alpha: 0.5 will return three numbers, none of which is wrong arithmetic and only one of which means anything.

Applies to — net10.0, netstandard2.0.

See also — PinballLoss.Score, MeanAbsoluteError.PerOutput, the Python equivalence table.