English

Nearly Minimax-Optimal Rates for Noisy Sparse Phase Retrieval via Early-Stopped Mirror Descent

Signal Processing 2021-05-11 v1 Machine Learning Machine Learning

Abstract

This paper studies early-stopped mirror descent applied to noisy sparse phase retrieval, which is the problem of recovering a kk-sparse signal xRn\mathbf{x}^\star\in\mathbb{R}^n from a set of quadratic Gaussian measurements corrupted by sub-exponential noise. We consider the (non-convex) unregularized empirical risk minimization problem and show that early-stopped mirror descent, when equipped with the hyperbolic entropy mirror map and proper initialization, achieves a nearly minimax-optimal rate of convergence, provided the sample size is at least of order k2k^2 (modulo logarithmic term) and the minimum (in modulus) non-zero entry of the signal is on the order of x2/k\|\mathbf{x}^\star\|_2/\sqrt{k}. Our theory leads to a simple algorithm that does not rely on explicit regularization or thresholding steps to promote sparsity. More generally, our results establish a connection between mirror descent and sparsity in the non-convex problem of noisy sparse phase retrieval, adding to the literature on early stopping that has mostly focused on non-sparse, Euclidean, and convex settings via gradient descent. Our proof combines a potential-based analysis of mirror descent with a quantitative control on a variational coherence property that we establish along the path of mirror descent, up to a prescribed stopping time.

Keywords

Cite

@article{arxiv.2105.03678,
  title  = {Nearly Minimax-Optimal Rates for Noisy Sparse Phase Retrieval via Early-Stopped Mirror Descent},
  author = {Fan Wu and Patrick Rebeschini},
  journal= {arXiv preprint arXiv:2105.03678},
  year   = {2021}
}

Comments

arXiv admin note: text overlap with arXiv:2010.10168