基于生成式人工智能的算力网络自智优化研究综述

摘要：生成式人工智能（GAI）技术可以在多样化业务处理过程中赋予算力网络精准的意图分析能力，通过算网专家库的构建进而辅助算力网络实现高效的自适应智能决策，通过模型微调技术使资源配置决策适应突发网络变化，为用户提供精准且稳定的服务。基于上述目标，首先介绍生成式人工智能和算力网络概述，然后讨论了基于生成式人工智能的网络自智优化相关研究进展，创新性提出生成式算力网络的架构，对其核心流程和所需关键技术进行讨论，并对所提架构的优越性进行仿真验证和分析，最后对生成式算力网络应用场景进行分析，期望对该领域的后续研究提出可供借鉴的新思路。

关键词：生成式人工智能；算力网络；意图分析；模型微调

Abstract: Generative artificial intelligence (GAI) technologies can  endow the computing power networks (CPNs) with precise intent analysis capabilities in diverse business processing scenarios. By constructing an expert database within the CPNs, it assists in achieving efficient adaptive intelligent decision-making. Through model fine-tuning techniques, the resource allocation decisions can adapt to sudden network changes, providing users with accurate and stable services. Based on these objectives, this paper firstly introduces an overview of GAI and CPNs, then discusses the research progress on network self intelligence optimization based on GAI. A novel architecture for generative computing power networks is proposed, along with discussions on its core processes and necessary key technologies. Furthermore, the superiority of the proposed architecture is validated and analyzed through simulation. Finally, an analysis of the application scenarios of generative computing power networks is provided, aiming to propose new perspectives for subsequent research in this field.

Keywords: generative artificial intelligence; computing power network; intentional analysis; model fine-tuning