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 scaling

Folds another batch into the largest absolute value and the scale, as partial_fit does.

public MaxAbsScaler PartialFit(ReadOnlySpan<double> samples)

Parameters — samples is the next batch, row-major, with FeatureCount values per row.

Returns — a new scaler summarising every batch seen so far; the one it was called on is unchanged.

Exceptions — ArgumentException when samples holds no row, a partial one, or a non-finite value.

Example — the running maximum absolute value.

using Lodestar.Preprocessing;

MaxAbsScaler scaler = MaxAbsScaler.Fit([2.0, 1.0], featureCount: 1).PartialFit([-8.0]);

double largest = scaler.MaximumAbsolute[0];  // => 8
int seen = scaler.SampleCount;               // => 3

Remarks — the largest absolute value, so a negative batch can move it where a maximum would not. The near-constant floor is applied to the result, as Fit applies it.

It returns a new scaler and leaves this one alone — see StandardScaler.PartialFit for why.

Applies to — net10.0, netstandard2.0.

See also — MaxAbsScaler, MaxAbsScaler.Fit.