I'd suggest trying t-SNE [1] instead; you'll be losing almost all of the variance by projecting onto the first two eigenvectors produced by PCA. [1] http://kar

I'd suggest trying t-SNE [1] instead; you'll be losing almost all of the variance by projecting onto the first two eigenvectors produced by PCA. [1] http://karpathy.github.io/2014/07/02/visualizing-top-tweeps-with-t-sne-in-Javascript/ http://karpathy.github.io/2014/07/02/visualizing-top-tweeps-...