Linkblog: A Surprise For Big-Data Analytics (Gödel's Lost Letter)
Gödel’s Lost Letter blogs about a proof by Landweber, Lazar and Patel that might have some implications for data science.
tl;dr dimension reduction with continuous mappings is problematic. For any continuous map , , diameters of fibers of are not bounded.
(Fiber = preimage of a singleton.)