Real-world uses

Created for this library

1. 1.

   A recommendation team uses negative sampling during training so the softmax over millions of items is approximated by a manageable subset.

2. 2.

   An e-commerce ranking team uses negative sampling to construct training pairs efficiently when the catalog is too large for full softmax.

3. 3.

   A search team uses negative sampling to train its retrieval model efficiently without computing softmax over the full index.