SpaceMAP

Visualizing high-dimensional data by space expansion (ICML 2022)

SpaceMAP: Visualizing High-Dimensional Data by Space Expansion — Xinrui Zu, Qian Tao. ICML 2022. [Paper] · [Code] · [Video]

We propose SpaceMAP, a dimensionality-reduction method that visualizes data of any dimensionality on a 2-D map. Unlike previous methods, we analytically derive a transformation of distance between high- and low-dimensional spaces to match their capacity, and show that it provably reduces the intrinsic dimension of high-dimensional data within the maximum-likelihood intrinsic-dimensionality framework.

SpaceMAP visualization results compared with other dimensionality-reduction methods.
Result comparison with other dimensionality-reduction methods.
SpaceMAP optimization unrolling the Swiss Roll dataset.
The SpaceMAP optimization process as the “unrolling” of the Swiss Roll dataset.

References