Hacker News Reader

Parent Full thread

Galanwe

I don't get it as well.

The color dimension is configurable, so it makes sense.

But I don't get what the 3 dimensions of spatial coordinates are, and how to change them.

--- Edit

Alright my bad it's actually explained in the "tour": https://pair-code.github.io/understanding-umap/

It seems to be some kind of multivariate PCA.

View on HN

> It seems to be some kind of multivariate PCA.

No, UMAP is nonlinear. The general idea is that you generate a neighborhood graph of your data points, do a spectral embedding on that to get your initial result, and then do gradient descent to make its neighborhood graph closer to the high-dimensional one.

Reply on news.ycombinator.com