Some Theoretical Limitations of t-SNE

TL;DR

This paper provides a mathematical framework to understand the limitations of t-SNE in preserving important data features during dimension reduction.

Contribution

It establishes theoretical results demonstrating how t-SNE can lose significant data features in various scenarios.

Findings

t-SNE can significantly distort data features in certain cases

Mathematical analysis reveals specific limitations of t-SNE

Results help inform when t-SNE may not be suitable for data visualization

Abstract

t-SNE has gained popularity as a dimension reduction technique, especially for visualizing data. It is well-known that all dimension reduction techniques may lose important features of the data. We provide a mathematical framework for understanding this loss for t-SNE by establishing a number of results in different scenarios showing how important features of data are lost by using t-SNE.