In the context of single-cell analysis, subsampling (also referred to as downsampling) refers to taking a portion of the data based on some rule which often involves a random element to it. Subsampling is often employed to reduce the runtime or memory requirements of an algorithm. SPADE and t-SNE are two notable examples where subsampling is often employed.
Astrolabe does not employ subsampling when clustering the data or when running statistical analyses. Since the platform is fully scalable over Amazon Web Services, neither runtime nor memory are constraints for our current implementation.