Isolating the hard core of phaseless inference: the Phase selection formulation

TL;DR

This paper introduces a two-level 'Phase selection' formulation for real-valued phase retrieval, analytically characterizing the problem's free-energy landscape and demonstrating how meta-heuristics can achieve near-optimal inference performance.

Contribution

It proposes and analyzes a novel two-level formulation for phase retrieval, providing insights into the free-energy landscape and the effects of regularization and annealing on inference.

Findings

Identification of two free-energy branches near the Bayes threshold.

Regularization can lower the dataset size needed for accurate recovery.

Meta-heuristics with annealing approach Bayes-optimal efficiency.

Abstract

Real-valued Phase retrieval is a non-convex continuous inference problem, where a high-dimensional signal is to be reconstructed from a dataset of signless linear measurements. Focusing on the noiseless case, we aim to disentangle the two distinct sub-tasks entailed in the Phase retrieval problem: the hard combinatorial problem of retrieving the missing signs of the measurements, and the nested convex problem of regressing the input-output observations to recover the hidden signal. To this end, we introduce and analytically characterize a two-level formulation of the problem, called ``Phase selection''. Within the Replica Theory framework, we perform a large deviation analysis to characterize the minimum mean squared error achievable with different guesses for the hidden signs. Moreover, we study the free-energy landscape of the problem when both levels are optimized simultaneously, as a function of the dataset size. At low temperatures, in proximity to the Bayes-optimal threshold -- previously derived in the context of Phase retrieval -- we detect the coexistence of two free-energy branches, one connected to the random initialization condition and a second to the signal. We derive the phase diagram for a first-order transition after which the two branches merge. Interestingly, introducing an $L_2$ regularization in the regression sub-task can anticipate the transition to lower dataset sizes, at the cost of a bias in the signal reconstructions which can be removed by annealing the regularization intensity. Finally, we study the inference performance of three meta-heuristics in the context of Phase selection: Simulated Annealing, Approximate Message Passing, and Langevin Dynamics on the continuous relaxation of the sign variables. With simultaneous annealing of the temperature and the $L_2$ regularization, they are shown to approach the Bayes-optimal sample efficiency.