Dimensionality reduction · UMAP

Interactive UMAP simulator

Step by step: nearest neighbors, fuzzy union, embedding initialization and gradient-descent optimization. Each phase is narrated and shown side by side.

simulatorUMAP step by step
step 0 / 1
dataset
k = 5
min_dist = 0.10
original space
embedding 2Dwaiting for initialization...
state
phaseinit
k5
min_dist0.10
σ̄—
|edges|—
epoch0 / 200
loss—

Dataset with 250 points ready. Press next to find the 5 nearest neighbors of each point.