LaplacianEigenmaps<T>
Models & Types · AiDotNet.Preprocessing.DimensionalityReduction
Laplacian Eigenmaps for nonlinear dimensionality reduction.
Inheritance: TransformerBase<T, Matrix<T>, Matrix<T>> → LaplacianEigenmaps<T>
For Beginners
Laplacian Eigenmaps finds a low-dimensional representation where:
- Connected points in the graph stay close together
- The embedding respects the local geometry of the data
- It’s similar to spectral clustering but for dimensionality reduction
Use cases:
- Manifold learning when data lies on a curved surface
- Image segmentation and clustering
- When you want to preserve local connectivity
How It Works
Laplacian Eigenmaps constructs a weighted graph from the data and finds a low-dimensional embedding that preserves local neighborhood relationships by minimizing a cost function based on the graph Laplacian.
The algorithm:
- Constructs a k-nearest neighbor or epsilon-neighborhood graph
- Computes edge weights using a kernel (e.g., heat kernel)
- Computes the graph Laplacian: L = D - W
- Finds eigenvectors of the generalized eigenvalue problem: Ly = λD*y
Constructors
| Constructor | Summary |
|---|---|
LaplacianEigenmaps(int nComponents, int nNeighbors, double? radius, LaplacianAffinityType affinity, double gamma, int? randomState, int[] columnIndices) |
Creates a new instance of LaplacianEigenmaps. |
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. |