基于深度学习特征迁移的装备体系效能预测

发布时间：2018-03-25 04:10

本文选题：深度学习　切入点：迁移学习　出处：《系统工程与电子技术》2017年12期

【摘要】：针对武器装备体系效能评估在高维噪声小样本数据条件下准确性不高的问题,提出一种基于堆栈降噪自编码与支持向量回归机的混合模型。利用堆栈自编码神经网络对通用深层特征的自主抽取能力,通过在相似源域大数据上预训练混合模型,获得两任务间的共有特征知识,借助对该知识的迁移,在目标域微调该混合模型,从而提升支持向量回归机在小样本噪声数据上的学习预测精度。在一定作战想定背景下,结合武器装备体系仿真试验数据,对该混合模型进行验证。实验结果表明,与传统支持向量回归机等模型相比,所提模型能够更准确地评估装备效能。
[Abstract]:Aiming at the problem that the accuracy of weapon system effectiveness evaluation is not high under the condition of high dimension noise and small sample data, This paper presents a hybrid model based on stack denoising self-coding and support vector regression machine. By using the self-coding neural network of stack self-coding to extract general deep features, the hybrid model is pretrained on big data in the similar source domain. The common feature knowledge between the two tasks is obtained, and the hybrid model is fine-tuned in the target domain with the help of the migration of the knowledge, so as to improve the learning and prediction accuracy of the support vector regression machine on the small sample noise data. Combined with the simulation data of weapon equipment system, the hybrid model is verified. The experimental results show that compared with the traditional support vector regression model, the proposed model can evaluate the equipment effectiveness more accurately.
【作者单位】：国防大学信息作战与指挥训练教研部;航天飞行器生存技术与效能评估实验室;
【基金】：国家自然科学基金(61403401) 军民共用重大研究计划联合基金项目(U1435218)资助课题
【分类号】：E92;TP181

【相似文献】