Preprocessing onehotencoder transformsparse

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

Home › Preprocessing › Encoding and imputation

OneHotEncoder.TransformSparse

Encodes a row-major matrix of categories into a sparse matrix, OneHotEncoder(sparse_output=True)'s output and the reference's default.

public CsrMatrix TransformSparse(ReadOnlySpan<T> values)

Parameters — values is the categories to encode, row-major, with FeatureCount per row.

Returns — a rows × EncodedFeatureCount CsrMatrix holding exactly the ones Transform sets: at most one per feature and row, in ascending column order.

Exceptions — ArgumentException as for Transform.

Example — an infrequent category and an unknown one land in the same column.

using Lodestar.Abstractions;
using Lodestar.Preprocessing;

var options = new OneHotEncoderOptions { MinFrequency = 2, Unknown = UnknownCategory.Infrequent };
OneHotEncoder<string> encoder = Encoders.OneHot(["a", "a", "a", "b", "b", "c", "d"], 1, options);

CsrMatrix encoded = encoder.TransformSparse(["a", "c", "zzz"]);

int stored = encoded.NonZeroCount;                          // => 3
string columns = string.Join(",", encoded.ColumnIndices);   // => 0,2,2

Remarks — a dropped category and an ignored unknown store nothing, as sparse_output=True stores nothing for them. A high-cardinality column is where this matters: the dense encoding is mostly zeros, and every sparse consumer in these packages — the decompositions, the sparse scalers — takes a CsrMatrix.

Applies to — net10.0, netstandard2.0.

See also — OneHotEncoder.Transform, OneHotEncoder.