深度学习:多层神经网络的复兴与变革

 人工智能（AI）已经进入一个新的蓬勃发展期。推动这一轮AI 狂澜的是三大引擎，即深度学习（DL）、大数据和大规模并行计算，其中又以DL 为核心。本文回顾本轮“深度神经网络复兴”的基本情况，概要介绍常用的4 种深度模型，即：深度信念网络（DBN）、深度自编码网络（DAN）、深度卷积神经网络（DCNN）及长短期记忆递归神经网络（LSTM-RNN）。简要介绍深度学习在语音识别和计算机视觉领域几个重要任务上的应用效果情况。为便于应用DL，介绍了几种常用的深度学习开源平台。对深度学习带来的启示和变革做了一些开放式的评述，讨论了该领域的开放问题和发展趋势。

Deep learning: The revival and transformation of multi layer neural networks

Artificial intelligence (AI) has entered a new period of vigorous development. This round of AI topsy is driven by three engines, namely the depth of learning (DL), big data and massively parallel computing, with DL as the core. This article reviews from a historical perspective the basic situation of the round "deep neural networks renaissance", then summarizes the four common depth models: deep belief network (DBN), depth from network coding (DAN), deep convolutional neural networks (DCNN) and long short term memory recurrent neural network LSTM-RNN. After that, this paper briefly introduces the application effects of deep learning in speech recognition and computer vision. In order to facilitate the application of DL, it also introduces several commonly used deep learning platforms. Finally, the enlightenment and reform of deep learning are commented, and the open problems and development trend in this field are discussed.