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

Reads each bin back as its centre.

public double[] InverseTransform(ReadOnlySpan<double> encoded)

Parameters — encoded is a matrix this discretizer produced, row-major, OutputFeatureCount values per row.

Returns — one value per input feature, per row: the centre of the bin each code names.

Exceptions — ArgumentException when encoded holds a partial row or a non-finite value; when an ordinal code, truncated toward zero, falls outside [-n, n) for a feature of n bins — the reference indexes with astype(np.int64), so a negative code counts from the last bin and one past either end raises IndexError; or when a one-hot block is all zeros, which the reference's encoder refuses. A one-hot block is read by its first largest value, numpy's argmax (#1230).

Example — the four bin codes read back.

using Lodestar.Preprocessing;

double[] ages = [19.0, 22.0, 25.0, 31.0, 38.0, 44.0, 52.0, 61.0, 67.0, 74.0];

KBinsDiscretizer bins = KBinsDiscretizer.Fit(
    ages, 1, new KBinsDiscretizerOptions { BinCount = 4, Encoding = BinEncoding.Ordinal });

double[] centres = bins.InverseTransform([0.0, 1.0, 2.0, 3.0]);

double firstBin = centres[0];    // => 22
double lastBin = centres[3];     // => 67.5

double roundTrip = bins.InverseTransform(bins.Transform([44.0]))[0];   // => 51

Remarks — this is not a round trip, and cannot be. The discretizer replaced a measurement with the group it belongs to, and the only thing left to answer with is where that group sits: 44 comes back as 51, the centre of the bin from 41 to 61. Use it to read a discretized column in the units it came from, not to recover the column.

The reference's own inverse does the same, and its documentation says so in the same words: the returned value is the bin's centre.

Applies to — net10.0, netstandard2.0.

See also — KBinsDiscretizer.Transform, KBinsDiscretizer.