HashingEncoder<T>

Models & Types · AiDotNet.Preprocessing.Encoders

Encodes categorical features using feature hashing (hashing trick).

Inheritance: TransformerBase<T, Matrix<T>, Matrix<T>> → HashingEncoder<T>

For Beginners

Instead of creating one column per category:

  • Hash encoding creates a fixed number of columns (e.g., 8)
  • Each category is hashed to one of these columns
  • Multiple categories may share the same column (collision)

Pros: Fixed memory, handles new categories, fast Cons: Information loss from collisions, not reversible

How It Works

HashingEncoder uses a hash function to map categories to a fixed number of columns. This is useful for high-cardinality categorical features where one-hot encoding would create too many columns.

Unlike other encoders, HashingEncoder doesn’t need to store the category mappings, making it memory-efficient and able to handle previously unseen categories.

Constructors

Constructor Summary
HashingEncoder(int nComponents, bool alternateSign, int[] columnIndices) Creates a new instance of HashingEncoder.

Properties

Property Summary
bool AlternateSign { get; } Gets whether alternate signs are used for hash collisions.
bool SupportsInverseTransform { get; } Gets whether this transformer supports inverse transformation.
int NComponents { get; } Gets the number of hash components (output features per encoded column).

Methods

Method Summary
string[] GetFeatureNamesOut(string[] inputFeatureNames) Gets the output feature names after transformation.