Unifying AMP Algorithms for Rotationally-Invariant Models
Statistics Theory
2024-12-03 v1 Information Theory
Machine Learning
math.IT
Probability
Statistics Theory
Abstract
This paper presents a unified framework for constructing Approximate Message Passing (AMP) algorithms for rotationally-invariant models. By employing a general iterative algorithm template and reducing it to long-memory Orthogonal AMP (OAMP), we systematically derive the correct Onsager terms of AMP algorithms. This approach allows us to rederive an AMP algorithm introduced by Fan and Opper et al., while shedding new light on the role of free cumulants of the spectral law. The free cumulants arise naturally from a recursive centering operation, potentially of independent interest beyond the scope of AMP. To illustrate the flexibility of our framework, we introduce two novel AMP variants and apply them to estimation in spiked models.
Cite
@article{arxiv.2412.01574,
title = {Unifying AMP Algorithms for Rotationally-Invariant Models},
author = {Songbin Liu and Junjie Ma},
journal= {arXiv preprint arXiv:2412.01574},
year = {2024}
}