Dataset documentation and intended uses. Recommended documentation frameworks include datasheets for datasets, dataset nutrition labels, data statements for NLP, and accountability frameworks. Author statement that they bear all responsibility in case of violation of rights, etc., and Links to access the dataset and its metadata. Simulation environments should link to (open source) code repositories. The dataset itself should ideally use an open and widely used data format.

We also evaluate our algorithm empirically. By leveraging a state-of-the-art nearest-neighbor search library, we obtain a fast and accurate Centroid-Linkage HAC algorithm.

Each step replaces the two mostsimilar clusters by its union.

FCVC-Aircraft is released "for non-commercial research purposes only". Stanfard Cars uses the same license as Imagenet. Describable Textures Dataset is released for research purpose only. Flowers 102 is licensed under the GNUGeneral Public License, version 2. Food-101 was curated by [9] under fair use of data from foodsplotting.com, which has ceased operation in 2018. The datasets we use were provided with consent by their original curators and do not contain personally identifiable information toourknowledge,andwedonotintroduce anydatacurationin ourwork.

We also evaluate our algorithm empirically. By leveraging a state-of-the-art nearest-neighbor search library, we obtain a fast and accurate Centroid-Linkage HAC algorithm.