Metrics rootmeansquaredlogerror peroutput

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

RootMeanSquaredLogError.PerOutput

One root mean squared log error per output, unreduced.

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

Parameters — yTrue and yPred are the true and predicted values, row-major, every value above −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, it holds a non-finite value, or either array holds a value at or below −1; ArgumentOutOfRangeException when outputCount is below one.

Example — the same four counts read as two samples of two outputs.

using Lodestar.Metrics;

double[] yTrue = [3.0, 5.0, 2.5, 7.0];
double[] yPred = [2.5, 5.0, 4.0, 8.0];

double[] perOutput = RootMeanSquaredLogError.PerOutput(yTrue, yPred, outputCount: 2);
double first = perOutput[0];    // => 0.2693…
double second = perOutput[1];   // => 0.0832…

Remarks — the entries are the square roots of MeanSquaredLogError.PerOutput's, and they are the array Score reduces. Every entry is in log units, so unlike RootMeanSquaredError.PerOutput these really are comparable across outputs even when the targets count different things — which is one of the better reasons to model counts in log space in the first place.

The trap is inherited whole from the squared form: one value at or below −1 anywhere refuses the call, per span and not per output, so a column that can legitimately go negative has to be scored separately.

Applies to — net10.0, netstandard2.0.

See also — RootMeanSquaredLogError.Score, MeanSquaredLogError.PerOutput, RootMeanSquaredError.PerOutput, the Python equivalence table.