Parameter Training Efficiency Aware Resource Allocation for AIGC in Space-Air-Ground Integrated Networks

TL;DR

This paper introduces PARA, a resource allocation method that enhances parameter training efficiency for AIGC in space-air-ground networks by jointly optimizing user association, offloading, and resources.

Contribution

It proposes a novel PARA technique that effectively solves a complex optimization problem for resource allocation in SAGIN with limited resources.

Findings

PARA outperforms baseline methods in simulations.

The approach effectively maximizes parameter training efficiency.

Solid proofs ensure the solution's optimality within constraints.

Abstract

With the evolution of artificial intelligence-generated content (AIGC) techniques and the development of space-air-ground integrated networks (SAGIN), there will be a growing opportunity to enhance more users' mobile experience with customized AIGC applications. This is made possible through the use of parameter-efficient fine-tuning (PEFT) training alongside mobile edge computing. In this paper, we formulate the optimization problem of maximizing the parameter training efficiency of the SAGIN system over wireless networks under limited resource constraints. We propose the Parameter training efficiency Aware Resource Allocation (PARA) technique to jointly optimize user association, data offloading, and communication and computational resource allocation. Solid proofs are presented to solve this difficult sum of ratios problem based on quadratically constrained quadratic programming (QCQP), semidefinite programming (SDP), graph theory, and fractional programming (FP) techniques. Our proposed PARA technique is effective in finding a stationary point of this non-convex problem. The simulation results demonstrate that the proposed PARA method outperforms other baselines.