Nystrom method for out-of-sample embedding — nystrom

Approximate the embedding of a new data point using the Nystrom method, which is particularly useful for large datasets and data-dependent embedding spaces, such as multidimensional scaling (MDS).

Usage

nystrom_embedding(
  new_data,
  landmark_data,
  kernel_function,
  eigenvectors,
  eigenvalues,
  ...
)

Arguments

new_data: A matrix or data frame containing the new data points to be projected.
landmark_data: A matrix or data frame containing the landmark data points used for approximation.
kernel_function: A function used to compute the kernel matrix (e.g., a distance function for MDS).
eigenvectors: A matrix containing the eigenvectors obtained from the eigendecomposition of the kernel matrix between the landmark points.
eigenvalues: A vector containing the eigenvalues obtained from the eigendecomposition of the kernel matrix between the landmark points.
...: Additional arguments passed to the kernel_function.

Value

A matrix containing the approximate embedding of the new_data in the data-dependent space.