spectral_clustering: Spectral clustering of an affinity matrix

In MosaiClusteR: An Umbrella Framework for Multi-Source and Multi-Omics Clustering

Spectral clustering of an affinity matrix

Description

Normalised (Ng-Jordan-Weiss) spectral clustering. Given a symmetric, non-negative affinity matrix it forms the normalised Laplacian, embeds the objects in the space of its leading eigenvectors and applies k-means.

Usage

spectral_clustering(affinity, k = NULL, max_k = 10, nstart = 10)

Arguments

affinity	A symmetric n \times n non-negative similarity/affinity matrix over the objects.
k	Number of clusters. If NULL, chosen by the largest eigengap of the normalised Laplacian (searched over 2:max_k).
max_k	Upper bound for the eigengap search when k = NULL.
nstart	Number of k-means restarts.

Value

A list with cluster (named integer labels), k and the spectral embedding.

References

Ng, A. Y., Jordan, M. I. & Weiss, Y. (2002). On spectral clustering: analysis and an algorithm. NIPS 14.

See Also

NEMO, SNF

Examples

set.seed(1)
X <- rbind(matrix(rnorm(20 * 5, 0), 20), matrix(rnorm(20 * 5, 6), 20))
A <- exp(-as.matrix(dist(X))^2 / 2)
spectral_clustering(A, k = 2)$cluster[1:5]

MosaiClusteR documentation built on Oct. 25, 2026, 1:08 a.m.