LTSA<T>

Models & Types ยท AiDotNet.Preprocessing.DimensionalityReduction

Local Tangent Space Alignment for nonlinear dimensionality reduction.

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

For Beginners

LTSA is an improved version of LLE that:

  • Uses tangent spaces (local linear approximations) at each point
  • Better handles points with different local geometries
  • More mathematically principled than standard LLE

Use cases:

  • When standard LLE produces poor results
  • Manifolds with varying curvature
  • When you need more stable embeddings

How It Works

LTSA is a manifold learning algorithm that uses local tangent spaces to compute a global embedding. It estimates the tangent space at each point using PCA on neighbors, then aligns these local coordinate systems.

The algorithm:

  1. Find k-nearest neighbors for each point
  2. Compute local tangent space via PCA on each neighborhood
  3. Compute local coordinates in tangent space
  4. Align tangent spaces by minimizing reconstruction error

Constructors

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

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.