1. Data preprocessing • quollr

Here, we’ll walk through the process of preprocessing 2-D embedding data to obtain regular hexagons.

library(quollr)
library(dplyr)

First, you’ll need 2-D embedding data generated for your high-dimensional data. For our example, we’ll use a $3\text{-}D$ S-curve dataset with four additional noise dimensions (scurve). We’ve used UMAP as our non-linear dimension reduction method (NLDR) to generate embeddings for the scurve data.

scaled_umap <- gen_scaled_data(nldr_data = scurve_umap)

glimpse(scaled_umap)
#> List of 3
#>  $ scaled_nldr: tibble [5,000 × 3] (S3: tbl_df/tbl/data.frame)
#>   ..$ emb1: num [1:5000] 0.707 0.231 0.232 0.79 0.761 ...
#>   ..$ emb2: num [1:5000] 0.839 0.401 0.215 0.564 0.551 ...
#>   ..$ ID  : int [1:5000] 1 2 3 4 5 6 7 8 9 10 ...
#>  $ lim1       : num [1:2] -14.4 13.3
#>  $ lim2       : num [1:2] -12.4 12.3

The function gen_scaled_data() standardises the 2-D embedding and rescales it so that hexagons generated during visualisation or analysis will be regular.