摘要
鉴于深度学习的研究和应用价值及在学术和工业领域中的重要地位,对目前有代表性的主流的深度学习网络模型进行介绍,概述了深度学习当前发展状态,综述了深度学习发展方向。首先介绍了深度学习的历史沿革,根据应用研究对四种主要深度学习网络进行介绍,然后从网络性能提升、网络体系发展、新学习模式探索、深度强化学习、可视化理论研究五个方面总结了目前深度学习的发展状态,最后提出下一步深度学习发展展望。可以看到:深度学习在不同领域都有广泛的应用,而且具有明显的优势,但也存在需要进一步深入探索的问题,如提高深度学习的智能性、提高无标签数据的利用率等。
Considering deep learning＇s value of researching and applying and the importance in academic and industry area,this paper reviews the main stream deep learning network models and gives theirs introduction.First,the history of deep learning is introduced.According to applying research,four deep learning network of the main streams are introduced.Second,the developing state of current deep learning is summarizedfrom five aspects which are network performance improvement,net system development,the new learning model to explore,deep reinforce learning and visualization theory research.Last,development prospect of deep learning comes up.Although deep learning outperform other methods abviously in different fields,there are still some issues needed to be solved,such as intelligence of deep learning improvement,raising the utilization ratio of data without labels.
出处
《舰船电子工程》
2017年第4期5-9,111,共6页
Ship Electronic Engineering
关键词
深度学习
卷积神经网络
半监督学习
深度强化学习
人工智能
deep learning
convolution neural network
semi-supervised learning
deep reinforce learning
artificial intelligence
