UMAP — uniform manifold approximation

Upload high-dimensional data (CSV or JSON with label + features) and generate a 2D embedding. View the final result or turn on step-by-step mode to follow the neighborhood graph construction and the optimization.

1. input data (high-dimensional)

Drop a CSV or JSON here

or click to choose

CSV JSON

CSV: first column is the label, the rest are numeric features. JSON: { labels: [...], features: [[...]] }.

UMAP — embedding 2DBatch mode: runs full UMAP in a Web Worker without blocking the UI.

step-by-step mode

k = 15

min_dist = 0.10

learning_rate = 1.00

epochs = 200

load a dataset to get started

No dataset loaded. Upload a CSV/JSON with at least one label column and two or more feature columns.