t-SNE Exaggerates Clusters, Provably

Noah Bergam, Szymon Snoeck, Nakul Verma

Keywords: nonlinear dimension reduction, data visualization, t-SNE

TL;DR: t-SNE visualizations can overemphasize clusters and suppress outliers significantly

Abstract: Central to the widespread use of t-distributed stochastic neighbor embedding (t-SNE) is the conviction that it produces visualizations whose structure roughly matches that of the input. To the contrary, we prove that (1) the strength of the input clustering, and (2) the extremity of outlier points, cannot be reliably inferred from the t-SNE output. We demonstrate the prevalence of these failure modes in practice as well.

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Submission Number: 8125