Predicting carbon and water vapor fluxes using machine learning and novel feature ranking algorithms

Xia Cui

doi:10.1016/j.scitotenv.2021.145130

Efficient demixing of multiplex graph signals: A Convex-Concave optimization approach

2026-05-23

0

PRE

AI

W

Wang, Yunhang

L

Liu, Zhaoting *

L

Li, Ran

H

Huang, Ying

DOI：10.1016/j.sigpro.2026.110612

Original

Original request for help

Save

Abstract

En 中文

This paper investigates the demixing of an observed signal, modeled as the sum of multiple signals residing on multiplex graphs. We propose two Graph Signal Blind Separation (GSBS) methodologies based on the smooth graph signal model. These methodologies decompose the observed signal into distinct, structured components, enabling the recovery of individual signals while preserving their intrinsic structures and inter-layer dependencies. The proposed GSBS methods reformulate the graph signal blind separation problem as nonsmooth constrained optimization tasks and employ a convex-concave saddle point optimization framework for efficient recovery of the original graph signals. The convergence properties of the proposed algorithms are analyzed, with a particular focus on the relationship between convergence and step size. Additionally, we introduce an enhanced version of the GSBS algorithms incorporating Anderson acceleration, which leverages fixed-point iteration to improve convergence rates. Experimental results demonstrate the effectiveness of our proposed methods, showcasing their superior performance in comparison to existing approaches.

Keywords:

Smooth graph signals

Blind separation

Signal recovery

Saddle point optimization

Efficient demixing of multiplex graph signals: A Convex-Concave optimization approach

Abstract

Journal

Organization

Cited Papers

Citing Papers