Deep Learning前世今生,系列转载

深度学习是 机器学习研究中的一个新的领域,其动机在于建立、模拟人脑进行分析学习的神经网络,它模仿人脑的机制来解释数据,例如图像,声音和文本。深度学习是 无监督学习的一种。
深度学习的概念源于 人工神经网络的研究。含多隐层的多层感知器就是一种深度学习结构。深度学习通过组合低层特征形成更加抽象的高层表示属性类别或特征,以发现数据的分布式特征表示。
深度学习的概念由Hinton等人于2006年提出。基于深信度网(DBN)提出非监督贪心逐层训练算法,为解决深层结构相关的优化难题带来希望,随后提出多层自动编码器深层结构。此外Lecun等人提出的卷积神经网络是第一个真正多层结构学习算法,它利用空间相对关系减少参数数目以提高训练性能。

一、Deep Learning的前世今生

图灵在 1950 年的论文里,提出图灵试验的设想,即,隔墙对话,你将不知道与你谈话的,是人还是电脑 [1]。

这无疑给计算机,尤其是人工智能,预设了一个很高的期望值。但是半个世纪过去了,人工智能的进展,远远没有达到图灵试验的标准。这不仅让多年翘首以待的人们,心灰意冷,认为人工智能是忽悠,相关领域是“伪科学”。

2008 年 6 月,“连线”杂志主编,Chris Anderson 发表文章,题目是 “理论的终极,数据的泛滥将让科学方法过时”。并且文中还引述经典著作 “人工智能的现代方法”的合著者,时任 Google 研究总监的 Peter Norvig 的言论,说 “一切模型都是错的。进而言之,抛弃它们,你就会成功”[2]。

言下之意,精巧的算法是无意义的。面对海量数据,即便只用简单的算法,也能得到出色的结果。与其钻研算法,不如研究云计算,处理大数据。

如果这番言论,发生在 2006 年以前,可能我不会强力反驳。但是自 2006 年以来,机器学习领域,取得了突破性的进展。

图灵试验,至少不是那么可望而不可即了。至于技术手段,不仅仅依赖于云计算对大数据的并行处理能力,而且依赖于算法。这个算法就是,Deep Learning。

借助于 Deep Learning 算法,人类终于找到了如何处理 “抽象概念”这个亘古难题的方法。

于是学界忙着延揽相关领域的大师。Alex Smola 加盟 CMU,就是这个背景下的插曲。悬念是 Geoffrey Hinton 和 Yoshua Bengio 这两位牛人,最后会加盟哪所大学。

Geoffrey Hinton 曾经转战 Cambridge、CMU,目前任教University of Toronto。相信挖他的名校一定不少。

Yoshua Bengio 经历比较简单,McGill University 获得博士后,去 MIT 追随 Mike Jordan 做博士后。目前任教 University of Montreal。

Deep Learning 引爆的这场革命,不仅学术意义巨大,而且离钱很近,实在太近了。如果把相关技术难题比喻成一座山,那么翻过这座山,山后就是特大露天金矿。技术难题解决以后,剩下的事情,就是动用资本和商业的强力手段,跑马圈地了。

于是各大公司重兵集结,虎视眈眈。Google 兵分两路,左路以 Jeff Dean 和 Andrew Ng 为首,重点突破 Deep Learning 等等算法和应用 [3]。

Jeff Dean 在 Google 诸位 Fellows 中,名列榜首,GFS 就是他的杰作。Andrew Ng 本科时,就读 CMU,后来去 MIT 追随 Mike Jordan。Mike Jordan 在 MIT 人缘不好,后来愤然出走 UC Berkeley。Andrew Ng 毫不犹豫追随导师,也去了 Berkeley。拿到博士后,任教 Stanford,是 Stanford 新生代教授中的佼佼者,同时兼职 Google。

Google 右路军由 Amit Singhal 领军,目标是构建 Knowledge Graph 基础设施。

1996 年 Amit Singhal 从 Cornell University 拿到博士学位后,去 Bell Lab 工作,2000 年加盟 Google。据说他去 Google 面试时,对 Google 创始人 Sergey Brian 说,“Your engine is excellent, but let me rewirte it!”[4]

换了别人,说不定一个大巴掌就扇过去了。但是 Sergey Brian 大人大量,不仅不怪罪小伙子的轻狂,反而真的让他从事新一代排名系统的研发。Amit Singhal 目前任职 Google 高级副总裁,掌管 Google 最核心的业务,搜索引擎。

Google 把王牌中之王牌,押宝在 Deep Learning 和 Knowledge Graph 上,目的是更快更大地夺取大数据革命的胜利果实。

Reference

[1] Turing Test.

http://en.wikipedia.org/wiki/Turing_test

[2] The End of Theory: The Data Deluge Makes the Scientific Method Obsolete

http://www.wired.com/science/discoveries/magazine/16-07/pb_theory

[3] Introduction to Deep Learning.

http://en.wikipedia.org/wiki/Deep_learning

[4] Interview with Amit Singhal, Google Fellow.

http://searchengineland.com/interview-with-amit-singhal-google-fellow-121342

原文链接:http://blog.sina.com.cn/s/blog_46d0a3930101fswl.html

作者微博:http://weibo.com/kandeng#1360336038853

  • 0
    点赞
  • 0
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
Deep Learning, Vol. 1: From Basics to Practice By 作者: Andrew Glassner Pub Date: 2018 ISBN: n/a Pages: (909 of 1750) Language: English Format: PDF People are using the tools of deep learning to change how we think about science, art, engineering, business, medicine, and even music. This book is for people who want to understand this field well enough to create deep learning systems, train them, and then use them with confidence to make their own contributions. The book takes a friendly, informal approach. Our goal is to make the ideas of this field simple and accessible to everyone, as shown in the Contents below. Since most practitioners today use one of several free, open-source deep-learning libraries to build their systems, the hard part isn’t in the programming. Rather, it’s knowing what tools to use, and when, and how. Building a working deep learning system requires making a series of technically informed choices, and with today’s tools, those choices require understanding what’s going on under the hood. This book is designed to give you that understanding. You’ll be able to choose the right kind of architecture, how to build a system that can learn, how to train it, and then how to use it to accomplish your goals. You’ll be able to read and understand the documentation for whatever library you’d like to use. And you’ll be able to follow exciting, on-going breakthroughs as they appear, because you’ll have the knowledge and vocabulary that let you read new material, and discuss it with other people doing deep learning. The book is extensively illustrated with over 1000 original figures. They are also all available for free download, for your own use. You don’t need any previous experience with machine learning or deep learning for this book. You don’t need to be a mathematician, because there’s nothing in the book harder than the occasional multiplication. You don’t need to choose a particular programming language, or library, or piece of hardware, because our approach is largely independent of those things. Our focus is on the principles and techniques that are applicable to any language, library, and hardware. Even so, practical programming is important. To stay focused, we gather our programming discussions into 3 chapters that show how to use two important and free Python libraries. Both chapters come with extensive Jupyter notebooks that contain all the code. Other chapters also offer notebooks for for every Python-generated figure. Our goal is to give you all the basics you need to understand deep learning, and then show how to use those ideas to construct your own systems. Everything is covered from the ground up, culminating in working systems illustrated with running code. The book is organized into two volumes. Volume 1 covers the basic ideas that support the field, and which form the core understanding for using these methods well. Volume 2 puts these principles into practice. Deep learning is fast becoming part of the intellectual toolkit used by scientists, artists, executives, doctors, musicians, and anyone else who wants to discover the information hiding in their data, paintings, business reports, test results, musical scores, and more.
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值