SSP 2018: Unsupervised Learning from Max Entropy to Deep Generative Networks