Using negative log-likelihoods of TensorFlow Distributions as Keras losses

Nothing here yet. This short post is supposed to summarize its supplementary notebook. For now, please go there instead.