Preprocessing polynomialfeatures featurenames

Development build. This page describes main, not a released package. The latest published Lodestar.Preprocessing is 0.1.0 — read its documentation.

Home › Preprocessing › Feature transforming

PolynomialFeatures.FeatureNames

Each term's name, in the order the columns come out.

public static string[] FeatureNames(int featureCount, PolynomialFeaturesOptions options = null)

Parameters — featureCount is how many values each input row carries. options chooses the degree, the interaction restriction and the bias; null takes the reference's defaults.

Returns — one name per output column, in the reference's own spelling: 1 for the bias, x0 for a feature, x0^2 for a power, x0 x1 for a product.

Exceptions — ArgumentOutOfRangeException when featureCount is not positive or the degree is negative. ArgumentException when the degree is 0 and the bias is off, which leaves no term.

Example — the default expansion of two features.

using Lodestar.Preprocessing;

string[] names = PolynomialFeatures.FeatureNames(featureCount: 2);

string joined = string.Join("|", names);   // => 1|x0|x1|x0^2|x0 x1|x1^2

Remarks — this is the only way to read a coefficient back. The expansion returns an array of doubles with nothing to say which column is which, and the order is not obvious past degree two: x0^2 x1 comes before x0 x1^2, because the terms are ordered by their feature indices and not by their powers. A caller that fits a model on the expansion and prints its coefficients should print these beside them.

The names are composed invariantly, so a French or an Arabic culture produces the same strings as an English one: a name here is an identifier to match on, not a number to read.

Applies to — net10.0, netstandard2.0.

See also — PolynomialFeatures.Transform, PolynomialFeatures.OutputFeatureCount, the Python equivalence table.