摘要
本文依据模型结构对深度学习进行了归纳和总结,描述了不同模型的结构和特点。首先介绍了深度学习的概念及意义,然后介绍了4种典型模型:卷积神经网络、深度信念网络、深度玻尔兹曼机和堆叠自动编码器,并对近3年深度学习在语音处理、计算机视觉、自然语言处理以及医疗应用等方面的应用现状进行介绍,最后对现有深度学习模型进行了总结,并且讨论了未来所面临的挑战。
Deep learning has recently received widespread attention. Using a model structure, this paper gives a summarization and analysis on deep learning by describing and reviewing the structure and characteristics of differ- ent models. The paper firstly introduces the concept and significance of deep learning, and then reviews four typical models: a convolutional neural network; deep belief networks; the deep Boltzmann machine; and an automatic stacking encoder. The paper then concludes by reviewing the applications of deep learning as regards cessing, computer vision, natural language processing, medical science, and other aspects. Finally, deep learning model is summarized and future challenges discussed. speech pro- the existing
作者
刘帅师
程曦
郭文燕
陈奇
LIU Shuaishi CHENG Xi GUO Wenyan CHEN Qi(College of Electrical and Electronic Engineering, Changchun University of Technology, Changchun 130000, China)
出处
《智能系统学报》
CSCD
北大核心
2016年第5期567-577,共11页
CAAI Transactions on Intelligent Systems
基金
吉林省科技厅青年科研基金项目(20140520065JH
20140520076JH)
长春工业大学科学研究发展基金自然科学计划项目(2010XN07)
关键词
深度学习
卷积神经网络
深度信念网络
深度玻尔兹曼机
堆叠自动编码器
deep learning
convolutional neural network
deep belief networks
deep Bohzmann machine
automat-ic stacking encoder