Embedding Dimension of Contrastive Learning and k-Nearest Neighbors

Neural Information Processing Systems 

We study the embedding dimension of distance comparison data in two settings: contrastive learning and k-nearest neighbors (k-NN).