摘要
针对指挥信息系统模拟训练数据缺乏的问题,提出一种基于改进型生成对抗网络的模拟数据生成算法,构建指挥信息系统履行使命任务支撑能力指标体系,同时提出指挥信息系统非结构化信息处理的映射方式和拟合度因子,修正损失函数,从而提高优化水平。仿真结果表明,所提算法能够生成与原始数据分布相似度较高的数据集,为指挥信息系统全元素的模拟训练提供数据支撑。
Aiming at the lack of simulation training data in command information system,a simulation data generation algorithm based on evolutional generative adversarial networks(EGAN-SDG)is put forword,which constructs the index system of command information system’s ability to fulfill missions and tasks.The mapping mode of unstructured information processing in command information system and the fitting factor to modify the loss function to improve the optimization performance are proposed.Simulation results show that the algorithm can generate data sets with high similarity distribution of the original data,and can provide data support for the simulation training of all elements of the command information system.
作者
田相轩
石志强
TIAN Xiangxuan;SHI Zhiqiang(Information and Communication Department, Army Armored Force Academy, Beijing 100072, China)
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2021年第1期163-170,共8页
Systems Engineering and Electronics
基金
陆军武器装备军内科研科学研究重点资助课题。
关键词
生成对抗网络
指挥信息系统
模拟数据集
generative adversarial network
command information system
simulation data set
