摘要
在深度学习大热的现今,人们创造出了许多足以颠覆以前对于机器的认知的程序,如击败李世石的Alpha GO以及之后青出于蓝的Alpha Go Zero,在网络春晚上大放异彩的钢琴机器人特奥等。这些都是以前的浅层网络难以达到的水准,因此越来越多的研究者投入到深层神经网络之中,使得其逐渐成为了目前深度学习的主要形式,本文所介绍的卷积神经网络便是其中的一种代表性的结构。文章主要先讲述了卷积神经网络的发展;之后了解其结构以及各个部分分别在其中起到了怎样的作用;再次,介绍卷积神经网络的改进方法和目前的几个改进网络;最后,文章会介绍一下卷积神经网络具体应用的领域并且在结尾提出目前这一领域所需要面对的需要解决的问题。
As deep learning gets popular nowadays, numerous programs have been created and change people’s ideas on the machine.For example, Alpha Go that had defeated Lee Se-dol, Alpha Go Zero which works better than Alpha Go, and pianist robot Teo that had shone in the Spring Festival Web Gala. These achievements are far beyond the shallow network before. Therefore, the deep neural network is studied by more and more researchers and has now become a major form of deep learning. The convolutional neural network(CNN) introduced in this paper is a representative structure of the deep neural network. The paper introduces CNN as follows: first, the evolution;second, its structure and functions of every part of it;third, methods for improving it and several modified networks applied currently;fourth, fields for its specific application and problems to be confronted and resolved.
出处
《信息通信》
2019年第2期39-43,共5页
Information & Communications
关键词
卷积神经网络
深度学习
深层神经网络
Convolutional Neural Network
Deep Learning
Deep Neural Network
作者简介
俞颂华(1995-),男,浙江宁波人,硕士研究生,研究方向语音信号识别、智能计算与神经网络。
