Random determinants, the elastic manifold, and landscape complexity beyond invariance

25.01.2022 17:55 - 18:55

Benjamin McKenna (NYU)

Abstract: The Kac-Rice formula allows one to study the complexity of high-dimensional Gaussian random functions (meaning asymptotic counts of critical points) via the determinants of large random matrices. We present new results on determinant asymptotics for non-invariant random matrices, and use them to compute the (annealed) complexity for several types of landscapes. We focus especially on the elastic manifold, a classical disordered elastic system studied for example by Fisher (1986) in fixed dimension and by Mézard and Parisi (1992) in the high-dimensional limit. We confirm recent formulas of Fyodorov and Le Doussal (2020) on the model in the Mézard-Parisi setting, identifying the boundary between simple and glassy phases. Joint work with Gérard Ben Arous and Paul Bourgade.

M. Beiglböck, N. Berestycki, L. Erdös, J. Maas, F. Toninelli
IST Austria, Central Building, Mondi 2 (I01.01.008)