您的位置：首页 > 论文页面
深度学习最新研究进展综述
发表时间：2015-03-31 浏览量：8286 下载量：5185
| 全部作者： | 王蕾，张宝昌 |
| 作者单位： | 北京航空航天大学自动化科学与电气工程学院；意大利理工学院模式分析与计算机视觉实验室 |
| 摘 要： | 深度学习作为机器学习领域的新兴技术，给人工智能及相关领域带来了生机与活力。首先，对深度学习的重要性、概念及特点进行了详细阐释，说明深度学习的价值及意义。然后对深度学习目前较成熟的2个典型模型——卷积神经网络（convolutional neural networks，CNNs）和自动编码器进行详细综述，并对其最新研究进展应用进行了概括。之后对深度学习中比较有潜力及实际意义的2个模型进行了介绍：多层核函数机（multilayer kernel machines，MKMs）及深度时空推理网模型（deep spatio-temporal inference network，DeSTIN），为深度学习模型的发展方向注入新鲜力量。最后指出深度学习目前存在的缺点，并对未来发展方向进行阐述。 |
| 关 键 词： | 信息处理技术；深度学习；综述；神经网络；网络结构；模型比较 |
| Title： | Review on deep learning |
| Author： | WANG Lei，ZHANG Baochang |
| Organization： | School of Automation Science and Electrical Engineering,Beihang University; Pattern Analysis and Computer Vision(PAVIS), Italian Institute of Technology |
| Abstract： | Deep learning, which is an emerging technology of machine learning field, has brought vitality and vigor to artificial intelligence and related fields. Firstly, this paper gives the detailed elaboration of deep learning about importance, concept and characters. Then we illustrate the value and significance of deep learning. Secondly, it describes two kinds of typical deep learning models which are relatively mature at present in detail: convolutional neural networks (CNNs) and sparse auto-encoder network. We also summarize the latest application of them. Thirdly, we introduce two deep learning models which relatively have the potential and practical significance: multilayer kernel machines (MKMs) and deep spatio-temporal inference network (DeSTIN),which inject fresh energy to deep learning model development. At last, it concludes the existing shortcomings and summarizes the development trend of deep learning model. |
| Key words： | information processing technology; deep learning; review; neural network; network structure; model comparison |
| 发表期数： | 2015年3月第6期 |
| 引用格式： | 王蕾，张宝昌. 深度学习最新研究进展综述[J]. 中国科技论文在线精品论文，2015，8（6）：510-517. |
0
评论数 0请您登录
暂无评论
下载全文
多维论文