laplacian_eigenmap: Laplacian eigenmaps

In kieranrcampbell/embeddr: Laplacian eigenmaps & principal curves for pseudotemporal ordering of single-cell RNA-seq data

Description

Construct a laplacian eigenmap embedding

Usage

laplacian_eigenmap(W, measure_type = c("unorm", "norm"), p = 2,
  eig_tol = 1e-10)

Arguments

W	The weighted graph adjacency matrix
measure_type	Type of laplacian eigenmap (norm for normalised, unorm otherwise)
p	Dimension of the embedded space, default is 2
eig_tol	The value below which an eigenvalue is deemed to be 0. Ideally these would all be .Machine$double.eps but numerical instabilities mean they are often larger. A good cutoff seems to be around 1e-10. If more than one zero eigenvalue is found but this is inconsistent with the number of connected clusters then the user will be warned to adjust eig_tol.

Value

A list containing two entries:

• embeddingA p by nrow(W) matrix with the laplacian eigenmaps embedding

• connected_componentsA vector of length nrow(W) of integers specifying to which connected component each cell belongs

Examples

## create synthetic weight matrix
W <- matrix(sample(0:1, 25, replace = TRUE), nrow=5)
diag(W) <- 0 ; W <- 0.5*(W + t(W))
le <- laplacian_eigenmap(W) # returns a list with entries embedding and connected_components

kieranrcampbell/embeddr documentation built on May 20, 2019, 9:24 a.m.