Preprocessing quantiletransformer fit

Development build. This page describes main, not a released package. The latest published Lodestar.Preprocessing is 0.2.0 — read its documentation.

Home › Preprocessing › Feature transforming

QuantileTransformer.Fit

Fits the quantiles of every feature.

public static QuantileTransformer Fit(ReadOnlySpan<double> samples, int featureCount, QuantileTransformerOptions options = null)

Parameters — samples is the matrix, row-major, featureCount values per row; the span is read, never modified. options chooses how many quantiles to fit and what they map onto; null takes the reference's defaults, a thousand quantiles onto the unit interval.

Returns — a fitted QuantileTransformer.

Exceptions — ArgumentOutOfRangeException when featureCount is not positive or the quantile count is below one. ArgumentException when samples holds no row, a partial one, or a non-finite value.

Example — five quantiles over ten rows, which is what References and Quantiles hold.

using Lodestar.Preprocessing;

double[] skew = [1.0, 1.0, 2.0, 3.0, 5.0, 8.0, 13.0, 21.0, 34.0, 100.0];

QuantileTransformer coarse = QuantileTransformer.Fit(
    skew, 1, new QuantileTransformerOptions { QuantileCount = 5 });

int levels = coarse.References.Count;          // => 5
double quarterLevel = coarse.References[1];    // => 0.25
double quarterValue = coarse.Quantiles[0][1];  // => 2.25
double medianValue = coarse.Quantiles[0][2];   // => 6.5

Remarks — the quantile count is clamped to the number of rows. Asking for a thousand over ten rows fits ten, because there are only ten order statistics to read; the reference clamps the same way and warns. Asking for fewer than there are rows is the useful direction: it is a piecewise-linear approximation of the distribution, cheaper to hold and smoother to invert.

The fit sorts each column once and reads the levels off it, so it costs features × n log n and holds features × QuantileCount doubles afterwards.

Applies to — net10.0, netstandard2.0.

See also — QuantileTransformer.Transform, QuantileTransformerOptions.