inverse transform visualization

[Hendekagon]

Here I have used UMAP's inverse transform to create a visualization of all points in a 2d UMAP projection of a 7d hypercube. The original data points (4096 uniform random samples from the hypercube) are overlaid. For each 2d point in the normalized projected space I averaged the inverse transforms of 8 random nearby points, then translated that 7d vector to colour by mixing 7 colours regularly spaced around the hue circle in HSB colour space. Each saturated colour therefore corresponds with a dimension (or combination thereof) in the 7d original space:

Can anyone help me interpret this visualization ? I think we're looking at UMAP's fuzzy simplicial set.

For comparison here's another visualization of the 7d hypercube with only 512 data samples and 1 sample per pixel

and here's a 5-simplex (512 samples not shown, 1 sample per pixel)