A Experimental setup

A.1 Datasets We use two standardized few-shot image classification datasets. We also use the test splits of the following four datasets, as defined by Triantafillou et al. [57]. CUB-200: CUB-200 was collected by Welinder et al. The test split contains 30 classes. A.2 Network architectures We train two of the most popular network architectures in few-shot learning literature. Episode difficulty is approximately normally distributed - density plots.