Fig. 1
From: Shape-aware stochastic neighbor embedding for robust data visualisations
Four synthetic data sets. a Data sampled from two Gaussians with equal covariance matrix but different means. The red and blue clusters contains 1000 and 50 points, respectively. b Data sampled uniformly along two non-overlapping 1D nonlinear curves with Gaussian noise added. Each cluster contains 400 points. c Data contains 6 clusters with 100 points each. Data are sampled from Gaussians with equal covariance. The cluster means are arranged in two major groups, each containing 3 sub-clusters. The green lines are included for clarity. The numbers next to the lines indicate the ED between the cluster means. d Schematic illustration of data set containing 1440 points sampled on piece-wise linear manifolds in 60 dimensional space with Gaussian noise added to all the 60 dimensions