Generative Autoencoding of Dropout Patterns

Shunta Maeda

We propose a generative model termed Deciphering Autoencoders. In this model, we assign a unique random dropout pattern to each data point in the training dataset and then train an…