Groups
Category
Embedding spaces map discrete things like words or products to dense vectors so that similar items are close together.
Low-rank approximation replaces a big matrix with one that has far fewer degrees of freedom while preserving most of its action.