PCA 2D projections and MVIB 2D stochastic encoding.

This file presents, for each dataset, the plots of the PCA 2D projections, as well as the plots of the mean of the MVIB 2D stochastic encodings. For the MVIB stochastic encodings , the depicted points represent the mean μ. The K dimension of the latent space has been set to 2 in order to allow a 2D visualisation of the encodings. For training MVIB, the JMVIB−T objective (Eq 8) has been optimised. For MVIB, five copies of the means plots are available, as they are obtained by training the model with five different independent training-test random splits. Both the PCA and the MVIB plots have been created starting from the default datasets collection.
(ZIP)