HessianLLE<T>

Models & Types ยท AiDotNet.Preprocessing.DimensionalityReduction

Hessian Locally Linear Embedding for nonlinear dimensionality reduction.

Inheritance: TransformerBase<T, Matrix<T>, Matrix<T>> โ†’ HessianLLE<T>

For Beginners

Hessian LLE improves on LLE by:

  • Using curvature information (second derivatives)
  • Producing more faithful global embeddings
  • Better handling of manifolds with varying curvature

Use cases:

  • When standard LLE produces distorted embeddings
  • Manifolds with non-uniform curvature
  • When you need more accurate distance preservation

How It Works

Hessian LLE is an improvement over standard LLE that estimates the Hessian (second derivative) of the embedding function. It uses a quadratic form to measure local curvature and produces more globally coherent embeddings.

The algorithm:

  1. Find k-nearest neighbors for each point
  2. Compute local Hessian estimator using quadratic polynomials
  3. Build global Hessian matrix
  4. Find embedding by minimizing Hessian-based cost function

Constructors

Constructor Summary
HessianLLE(int nComponents, int nNeighbors, int? randomState, int[] columnIndices) Creates a new instance of HessianLLE.

Properties

Property Summary
Matrix<T> Embedding { get; } Gets the embedding result.
bool SupportsInverseTransform { get; } Gets whether this transformer supports inverse transformation.
int NComponents { get; } Gets the number of components (dimensions).
int NNeighbors { get; } Gets the number of neighbors.

Methods

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