Efficient power allocation using graph neural networks and deep algorithm unfolding

TL;DR

This paper introduces UWMMSE, a hybrid neural network approach using graph neural networks and algorithm unfolding to optimize power allocation in wireless networks, achieving near-WMMSE performance with lower complexity.

Contribution

The paper presents a novel hybrid neural architecture that combines algorithm unfolding with graph neural networks for efficient power allocation in wireless networks.

Findings

UWMMSE matches WMMSE performance

Significantly reduces computational complexity

Demonstrates robustness across different network sizes

Abstract

We study the problem of optimal power allocation in a single-hop ad hoc wireless network. In solving this problem, we propose a hybrid neural architecture inspired by the algorithmic unfolding of the iterative weighted minimum mean squared error (WMMSE) method, that we denote as unfolded WMMSE (UWMMSE). The learnable weights within UWMMSE are parameterized using graph neural networks (GNNs), where the time-varying underlying graphs are given by the fading interference coefficients in the wireless network. These GNNs are trained through a gradient descent approach based on multiple instances of the power allocation problem. Once trained, UWMMSE achieves performance comparable to that of WMMSE while significantly reducing the computational complexity. This phenomenon is illustrated through numerical experiments along with the robustness and generalization to wireless networks of different densities and sizes.