Tree-SNE: Hierarchical Clustering and Visualization Using t-SNE

  • 2020-02-13 18:11:00
  • Isaac Robinson, Emma Pierce-Hoffman
  • 47

Abstract

t-SNE and hierarchical clustering are popular methods of exploratory dataanalysis, particularly in biology. Building on recent advances in speeding upt-SNE and obtaining finer-grained structure, we combine the two to createtree-SNE, a hierarchical clustering and visualization algorithm based onstacked one-dimensional t-SNE embeddings. We also introduce alpha-clustering,which recommends the optimal cluster assignment, without foreknowledge of thenumber of clusters, based off of the cluster stability across multiple scales.We demonstrate the effectiveness of tree-SNE and alpha-clustering on images ofhandwritten digits, mass cytometry (CyTOF) data from blood cells, andsingle-cell RNA-sequencing (scRNA-seq) data from retinal cells. Furthermore, todemonstrate the validity of the visualization, we use alpha-clustering toobtain unsupervised clustering results competitive with the state of the art onseveral image data sets. Software is available athttps://github.com/isaacrob/treesne.

 

Quick Read (beta)

loading the full paper ...