Post by @molsi • Hey

this reminds me of an experiment we ran at ICLR'20, when we provided a calibrated score based on a simple Gaussian latent variable model with missing value