Abstract: With the development and progress of artificial intelligence, a large number of smart devices are used on the battlefield. In complex and ever-changing battlefield environments, these smart devices face challenges such as difficulty in recognition under complex military environments, limited computational resources on mobile devices, and poor interpretability of AI technologies. To address these issues, this paper proposes a practical solution for applying the theories, methods, and tools of network science to the design and analysis of artificial intelligence systems, and summarizes the research progress of network science empowering artificial intelligence at home and abroad and elaborates on the current research status of network science empowering artificial intelligence from three aspects: graph representation on the input and output end, graph representation on the model architecture end, and graph representation on the decision logic end. It also discusses the challenges faced by network science in empowering artificial intelligence and possible future development directions.