t-SNE is a cool technique, even if some of the parameters require manual judgment/tuning. I've actually been thinking about taking our various groupings and try

t-SNE is a cool technique, even if some of the parameters require manual judgment/tuning. I've actually been thinking about taking our various groupings and trying a combined analysis, maybe with multivariate distance matrix regression. One issue is that labels are not exclusive, some individual may belong to several.