深度学习与神经网络

全书分为7 个章节。第1 章绪论，梳理了人工智能不同技术流派的特点、深度学习的发展及前沿技术；第2 章介绍相关预备知识，包括线性代数、概率论、优化理论以及机器学习的基础知识；第3 章从前馈神经网络的基础模型——感知器出发，介绍前馈神经网络的基本结构以及涉及的激活函数、梯度下降、反向传播等内容；第4 章，介绍深度模型的优化问题，讨论了神经网络优化中常见的病态问题；第5 章介绍深度学习中的正则化方法，包括范数惩罚、数据集增强与噪声注入、提前停止等；第6 章介绍了卷积神经网络，以及卷积神经网络在计算机视觉领域的具体应用；第7 章通过实际案例介绍循环神经网络与卷积神经网络的结合应用。

赵金晶，女，1981年生，军事科学院系统工程研究院高级工程师，国防科技大学计算机学院博士毕业，主要研究方向为网络与信息安全、人工智能技术。先后承担国家自然科学基金、国家973计划、863计划等重大项目，曾获国家自然科学基金青年基金资助。获省部级科技进步奖二等奖5项、三等奖1项，发明专利20余项。发表学术论文80余篇，出版学术著作2部。电子邮箱：zhjj0420@126.com李虎，男，1987年生，军事科学院系统工程研究院工程师，国防科技大学计算机学院博士毕业，主要研究方向为网络与信息安全。先后承担、参与国家973计划、863计划、装备预研等各类科研项目10余项。获省部级科技进步奖二等奖2项、发明专利10余项，发表论文20余篇。电子邮箱：lihu@nudt.edu.cn张明，男，1990年生，军事科学院系统工程研究院工程师，北京系统工程研究所硕士毕业，主要研究方向为机器学习和人工智能安全。先后参与、主持国家自然科学基金、重点实验室基金、国家973重大项目、装备预研和国防科技创新特区等各类科研项目10余项。获省部级科技进步奖一等奖1项，二等奖2项。发表学术论文20余篇，其中SCI检索5篇，EI检索10余篇。电子邮箱：zm_stiss@163.com

第1 章 绪论····················································································.1 
1.1 人工智能·············································································.2 
1.1.1 人工智能技术的发展历程···············································.3 
1.1.2 人工智能技术的流派·····················································.9 
1.2 深度学习与神经网络概述······················································.11 
1.2.1 深度学习与神经网络技术的发展历程······························.11 
1.2.2 深度学习与神经网络的前沿技术····································.16 
1.3 深度学习系统架构·······························································.17 
1.4 深度学习框架·····································································.19 
1.5 深度学习的应用··································································.20 
1.5.1 计算机视觉·······························································.20 
1.5.2 语音语义··································································.21 
1.5.3 自然语言处理····························································.22 
1.6 人工智能潜在的安全风险······················································.22 
1.6.1 数据层面的风险·························································.23 
1.6.2 算法模型层面的风险···················································.23 
1.6.3 智能计算框架层面的风险·············································.23 
1.6.4 基础软硬件层面的风险················································.24 
1.6.5 应用服务层面的风险···················································.24 
本章小结·······&midd