A Review of t-SNE
Sangwon Jung, Tristan Dagobert, Jean-Michel Morel, Gabriele Facciolo
⚠ This is a preprint. It may change before it is accepted for publication.

Abstract

High dimensional data is difficult to visualize. T-Distributed Stochastic Neighbor Embedding (t-SNE) is a popular technique for dimensionality reduction enabling a planar visualization of a dataset preserving as much as possible its metric. This paper explores the theoretical background of t-SNE and its accelerated version. A comparison of the performance of t-SNE on various datasets with different dimensions is also performed.

Download