Some theoretical limitations of t-SNE
In the authors' words
t-distributed Stochastic Neighborhood Embedding (t-SNE) has gained popularity as a dimension reduction technique, especially for visualizing data. It is well-known that dimension reduction may lose important data features. We provide a mathematical framework for understanding this loss for t-SNE, establishing several quantitative results across different scenarios that demonstrate how important structural information is lost.
Main resultThe abstract does not state a limitation.
Appeared: Saturday, November 7. Proceedings of the National Academy of Sciences. Peer-reviewed journal.