An implementation of multiple maps t-distributed stochastic neighbor embedding (t-SNE) in R. Multiple maps t-SNE is a method for projecting high-dimensional data into several low-dimensional maps such that metric space properties are better preserved than they would be by a single map. Multiple maps t-SNE with only one map is equivalent to standard t-SNE. When projecting onto more than one map, multiple maps t-SNE estimates a set of latent weights that allow each point to contribute to one or more maps depending on similarity relationships in the original data. This implementation is a port of the original Matlab library by Laurens van der Maaten.
July 15, 2018