Skip to contents

Subset Cluster Data

Source: R/misread-tsne.R
subset_clusters_data.Rd

One tiny gaussian cluster inside of a big cluster from "How to Use t-SNE Effectively".

Usage

subset_clusters_data(n, dim = 2, big_sdev = 50)

Arguments

n

Number of points per gaussian.

dim

Dimension of the gaussians.

big_sdev

Standard deviation of the bigger cluster, default 50. The smaller cluster has a standard deviation of 1.

Value

Data frame with coordinates in the X1, X2 ... Xdim columns, and color in the color column.

Details

Creates a dataset consisting of two gaussians with the same center, but with the first cluster having a standard deviation of 1, and the second having a standard deviation of big_sdev (default 50). Points are colored depending on which cluster they belong to (small cluster is dark powder blue, large is light orange).

References

http://distill.pub/2016/misread-tsne/

See also

Other distill functions: circle_data(), cube_data(), gaussian_data(), grid_data(), link_data(), long_cluster_data(), long_gaussian_data(), ortho_curve(), random_circle_cluster_data(), random_circle_data(), random_jump(), random_walk(), simplex_data(), three_clusters_data(), trefoil_data(), two_clusters_data(), two_different_clusters_data(), unlink_data()

Examples

df <- subset_clusters_data(n = 50, dim = 2)

# 10D example where the big cluster is only twice the standard deviation of
# the small cluster
df <- subset_clusters_data(n = 50, dim = 10, big_sdev = 2)