17. Z. Allen-Zhu, R. Gelashvili, S. Micali, N. Shavit, Proc. Natl. Acad. Sci. U.S.A. 111, 16872–16876 (2014).
Zhu et al. consider whether the physical properties of neural tissue are actually conducive to performing Johnson-Lindenstrauss (JL) calculations in the brain. For example, the fact that neurons are either inhibitory or excitatory has certain implications for the signs (+ or -) in a JL matrix.
They determine that it is indeed possible to construct JL matrices that accurately reflect the physical properties of the brain.