1. 目的及意义

（1）深度学习简介

深度学习是近年来计算机科学领域比较热门的研究方向，是机器学习的一个分支。

机器学习指机器能够利用大数据集进行自主学习，而不是通过硬性编码规则，这种学习类型基于现代计算机强大的处理能力，能够在短时间内处理大数据集；并且由于大数据时代的到来，人们收集信息的速度和规模变得很大，从而为机器学习提供了海量数据集。机器学习是人工智能的核心，是使计算机具有智能的根本有效途径。

深度学习的概念源于人工神经网络的研究。神经网络是构成深度学习系统的框架，它由大量彼此相连、概念化的人造神经元组成，分为输入层、隐藏层和输出层。普通神经网络只有一个隐藏层，含有多个隐藏层的人工神经网络便称为深度学习系统。

深度学习分为监督式学习和非监督式学习两种学习方法。监督式学习应用了包含输入值和期望输出值的标记数据集，而非监督式学习使用没有具体结构的数据集。从长期来说，非监督式学习绝对会变得更加重要；但面对现实，近十年以来所做的都是监督式学习，都侧重学习，都有标签。在未来，当我们真正搞明白一些东西以后，更加神奇的非监督学习的效果将会大大改善，从而成为主流。

当前深度学习领域最新，最具前景的想法是Lan Goodfellow等提出的生成对抗网络概念。

Introduction of Deep learning

Deep learning is a very popular research direction in computer science in recent years,It's a branch of machine learning.

Machine learning refers to the ability of machines to learn autonomously by using large data sets, rather than by hard coding rules,this type of learning is based on the powerful processing power of modern computers and can process large data sets in a short time.And thanks to the advent of the big data era, the speed and scale of information gathering has become great, which provides a huge data set for machine learning.Machine learning is the core of artificial intelligence, which is the fundamental and effective way to make computers intelligent.

The concept of deep learning originates from the research of artificial neural network. Neural network is the framework of deep learning system, which consists of a large number of interconnected and conceptualized artificial neurons, which are divided into input layer, hidden layer and output layer. The common neural network has only one hidden layer, and the artificial neural network with multiple hidden layers is called the deep learning system.

Deep learning is divided into supervised learning and non-supervised learning. Supervised learning uses a tag data set that contains input values and expected output values, while supervised learning uses data sets with no specific structure. In the long run, unsupervised learning will definitely become more important. But in the face of reality, what has been done in the past ten years is supervised learning, which focuses on learning and has labels. In the future, when we really understand something, the effect of the more magical non-supervised learning will be greatly improved and become the mainstream.

The latest and most promising idea in the field of deep learning is the generation of anti-network concepts proposed by Lan Goodfellow.