Gaussian Critical-Threshold Instability in Real Phase Retrieval

Christian E. H\"aggblom

In the authors' words

We establish the natural scale of instability in real Gaussian phase retrieval at the critical injectivity threshold. Let $A$ be a $(2 M - 1) \times M$ matrix with independent standard Gaussian entries and let $K_{M} = (M 2 M - 1)$ . For every $w_{M} \to \infty$ with $lo g w_{M} = o (M)$ , we prove that $P {(w_{M} M K_{M})^{- 1} \leq ω (A) \leq w_{M} / (M K_{M})} \to 1$ , where $ω (A)$ is the Balan--Wang stability parameter. Consequently, $- M^{- 1} lo g ω (A) \to lo g 4$ in probability. For full-spark matrices at this threshold, $ω (A)$ equals both the optimal lower Lipschitz constant of $x \mapsto ∣ A x ∣$ and the minimum least singular value over all square row submatrices. The upper bound follows from a second-moment analysis of overlapping minors. A weighted Gaussian inverse-tail asymptotic and inverse-Wishart concentration yield asymptotic independence for central overlaps; rectangular hard-edge bounds control the remaining overlaps. The matching lower bound follows from a union bound and a square Gaussian hard-edge estimate.

Main resultThe abstract does not state a limitation.

Gaussian Critical-Threshold Instability in Real Phase Retrieval

In the authors' words

More on Probability that day

Distributed Stochastic Approximation Algorithms and Heavy-Tailed Age of Information

Local Weak Limits for Equilibrium and Risk in Economic Networks

High-Dimensional Ultra-Log-Concave Distributions

Long-time behavior of McKean--Vlasov stochastic systems with singular coefficients

The Sharp R\'enyi and Tsallis Threshold in the Shepp--Olkin Concavity Problem