Implementing "Learning to Remember Rare Events" by Kaiser et. al in Pytorch. https://arxiv.org/pdf/1703.03129.pdf
Implementing a 2-layered CNN for classifying CIFAR-100 images. First we classify them by their superclass "coarse" labels in a CNN. Then we split the data according to this classification and train 20 separate CNNs to classify images according to their "fine" labels.