局域网中深度学习平台构建及应用

Construction and Application of Deep Learning Platform in LAN

随着飞行器研制领域各类设计、仿真和试验数据的日益增加,采用何种恰当的手段发掘海量数据的潜在价值成为了难题。深度学习作为当前人工智能领域备受瞩目的核心技术,在解决飞行器研制领域数据分析问题上或可带来突破。笔者以构建深度学习平台为出发点,介绍了飞行器研发局域网内典型操作系统Linux和Windows上Tensor Flow深度学习平台的构建方法,并阐述了基于该平台完成典型卷积神经网络模型Lenet-5的实现过程,一方面为同行提供参考,另一方面也为后续开展基于深度学习的流场、控制等学科分析打下基础。

As the increasing of design, simulation and experimental data of aircraft development, the right way to find the potential value of massive data is a difficult problem. Deep learning as the core technology of artificial intelligence may bring breakthroughs in solving the problem of aircraft development data analysis. The author focus on the construction of deep learning platform, present the method of build Tensor Flow platform on Linux and Windows operation systems in LAN, and also discussed the implementation of typical convolution neural network on the platform. On the one hand, the work will provide references for peers, On the other hand,it will lay a foundation for further analysis of flow field and control based on deep learning.