Lodestar.Preprocessing 0.2.0. This page is frozen at that release. Read the current documentation for what main says now. A link to a decision or a migration page follows main, and leaves the archive.

Home › Feature scaling

MaxAbsScalerOptions

Whether MaxAbsScaler clips a scaled value into [−1, 1].

public sealed record MaxAbsScalerOptions

Properties — Clip bounds Transform's output to [−1, 1], off by default as sklearn.preprocessing.MaxAbsScaler's clip is.

Example — the option exists for values the fit never saw.

using Lodestar.Preprocessing;

MaxAbsScaler scaler = MaxAbsScaler.Fit([1.0, 2.0], 1, new MaxAbsScalerOptions { Clip = true });

double bounded = scaler.Transform([8.0])[0];  // => 1

Remarks — one property rather than a range, because the range is not a choice here: dividing by the largest absolute value puts every fitted row inside [−1, 1] by construction, and clipping only decides what happens to a row that arrives later and is larger.

Applies to — net10.0, netstandard2.0.

See also — MaxAbsScaler, MaxAbsScaler.Fit.