Mapper

topological data analysis · in your browser

Reveal the shape of your data

Drop a high-dimensional CSV, or try a synthetic dataset on the left. The Mapper algorithm will distill its topology — loops, flares, branches — into an interactive 3D graph.

Drag rotate
Scroll zoom
Right-drag pan
Click focus a node

What's happening here

The Mapper algorithm builds a low-dimensional graph that preserves the topology of high-dimensional data. It works in four steps:

  1. Lens. Project every point through a scalar function — PCA, eccentricity, density, or a chosen column.
  2. Cover. Slice the lens range into overlapping intervals (resolution × overlap).
  3. Cluster. Inside each interval's preimage, run clustering in the original high-D space.
  4. Nerve. One node per cluster. Two clusters share an edge iff they share original data points.

Loops in the data become loops in the graph. Flares become flares. A "Y" appears wherever the data branches.

Reference: Singh, Mémoli & Carlsson (2007), Topological Methods for the Analysis of High Dimensional Data Sets.