overlay-centroids: Overlay centroids via k-means or graph-based clustering

In sa-lee/sneezy: Interactively explore non-linear projections like t-SNE

Description

Overlay centroids via k-means or graph-based clustering

Overylay k-means centroids onto an xy scatter

Usage

overlay_knn_centroids(x, y, indices, ...)

overlay_snn_centroids(x, y, indices, ...)

overlay_kmeans_centroids(x, y, num_centers, ...)

Arguments

x	numeric vectors to produce segments from
y	numeric vectors to produce segments from
indices	a neighborSet index matrix
...	additional arguments to pass to ggplot2::geom_segment
num_centers	the number of centroids to estimate

Details

The nearest neighbor centroids functions automatically estimate centroids from an indices matrix representing a nearest neighbor graph via a community detection algorithm (louvain) from the igraph package. The smaller the value of ncol(indices) the more centroids there will be. The kmeans method returns the centroids from run stats::kmeans with the user suppplied number of centers.

Value

a ggplot2::geom_point layer

sa-lee/sneezy documentation built on Nov. 25, 2019, 12:38 p.m.