Read "mathematical beauty"
In fact prepared to read "mathematical beauty" This book is from a very long time to start.
I remember last summer when it is read "the wave of the summit", then you know that Wu a person, very much like his writing style. So ready to read "mathematical beauty" of.
"Mathematical beauty," the book itself and "the wave of the summit" the same. Already we have a very good reputation in the number of readers, Dr. Wu's own scholarship and research, for many other people. Also formed a kind of psychological respect.
After I read the "wave of the summit," a kind of very heavy sense of history accumulation in the chest.
Have to admit that "the wave of the summit" For my influence.
"Mathematical beauty" in fact and "the wave of the summit" a very big difference, "the wave of the summit" can be said to be a more suitable for reading popular science class, there are authors ideas very deep in it; and "mathematical beauty" is relatively just for the reader to read this area.
Although Dr. Wu has "mathematical beauty" in a lot of theorems written in a very simple and obvious. But suppose for a liberal arts students, I think it is still very difficult to understand. Even for science background, I read in a year of "Higher Mathematics", a lot of detailed technical theorems which are the same as very strange. So prepare to read this book from last year to this year really read it. Separated by time quite a long time. And in between, I think there are a lot of chance and coincidence, I due to the work, understand some machine learning, artificial intelligence, contain knowledge Chinese sub-parts of speech, which is part of the knowledge, in the "mathematical beauty", the also it has a very good expression.
I think it is a hard work I had not sent for technical knowledge of some kind. I would like to start with the theoretical knowledge to start. First a general framework must understand, after a certain ability enough to do a partial block. This logic in a large company, it is basically impossible, a lot of systems. Containing tissue. Some people know the overall framework has also been very few, most people are busy with their own piece, finished both finished. As I learn Chinese word, and so on things.
After a certain basic knowledge, I would think to read "mathematical beauty" is a very happy thing, in simple terms, Dr. Wu to explain a lot of theoretical knowledge and narrative description. Very much time to make you feel a kind of enlightened.
I think Dr. Wu written word will always be a humanistic things inside. Feeling very mysterious, it can make you feel a sense of people feeling very good accomplishments in the academic heart. Like Wu narrative description of a lot of the same people. There are many very many successful people, a person is not tightly, or a very strange temperament, or a very unique ideas, but very many people have the same attitude and adhere to their own kind. In affect many other people. So they will be successful. Also led many other successful people.
After reading the "mathematical beauty", and it is very much a reflection of their own things. correct. Incorrect, there is.
Sometimes I think. Some universities Why so good. However, some universities are not so good. Very many people are willing to work very hard to go to a better school why, and attended the University of very many people should be aware that many other universities are self-study. You can be very teacher exchanges.
And good university I understand it, the reason why it is good, because it is first of all there are so many very good teacher concentrated there, so they can focus on many other concerns. Then attracted a lot of other good people, and these cattle also will lead good people. So it would form a cycle, due to good to good, change is good. This is a good development cycle.
Dr. Wu wrote in a postscript inside, I think it is good in particular:. "The world's best scholars always tell truths in simple terms the layman to listen to, rather than a simple trick of the complexity of the problem," Dr. Wu book also we talked about a lot of theory. In a lot of time, you may try to simplify it to go, simple direction to think about, perhaps the result will be very few, like the hidden Markov model, and so on. Or other scientific thought leaders.
After each book read Dr. Wu, always I think something more to say (although total. I have only read two).
Although the book is finished, but I always thought that inside, I can think of, there are many things to understand. I remember something that I am very impressed with the words "very many people reading, it should not pursue the number of school. And that you should read it. You used the time to think."
Here is what I used to put a book folder accompanied by:
Chapter 1 vs digital words and language and information
Words and language and mathematics, had had since generated from interlinked, although their development has been split, but finally can come together.
Chapter 2 NLP - from the rules to the statistical understanding of human-to-machine natural language understanding taking a big detour. Early research focused on Methods Jiyuguize, despite the capture of a few simple questions. But can not understand that fundamentally practical natural language. Until? Years later. People are experimenting with natural language processing based on the statistical methods used, have a breakthrough and practical products.
Chapter 3 statistical language model
Statistical language model is the basis of natural language processing, and is widely used in machine translation, speech recognition, printed or handwriting recognition, spelling correction, Chinese character input and query documents.
Chapter 4, talk about the Chinese word
Chinese word is the basis of Chinese information processing, which is the same taken a detour. Right now rely on statistical language model has basically overcome this problem.
Chapter 5 Hidden Markov Models
Hidden Markov model was originally used in the field of communications, and then extended to the speech and language processing, becoming a bridge between natural language processing and communications. The same time. Among the main tool of a hidden Markov model is machine learning.
Chapter 6 measure and the role of information
Information is able to quantify the measure. Entropy is not only a quantitative measure of information, but also the basis of the whole information theory.
It is for communication, data compression, natural language processing have a very strong guiding significance.
Chapter 7 Jared Nick and modern language processing
As the founder of modern natural language processing. Professor Jared Nick successfully applied mathematical principles of natural language processing field, his legendary life.
Chapter 8 simple beauty - Boolean algebra and Boolean algebra search engines index despite very easy. It is the foundation of computer science, it is not only logic and mathematics combined, and gave us a new perspective on the world, creating a digital age.
Chapter 9 graph theory and crawlers
Internet search engines need to actively download the entire page before indexing with a program on their own server, this program is called web crawlers, it's writing is based on the principle of Discrete Mathematics Graph Theory.
Chapter 10 PageRank - Google's Democratic vote-style screen name
Page Rank technology is an early killer. Its appearance makes a big step on the quality of web search. The principle behind it is a matrix computation graph theory and linear algebra.
Chapter 11 pages and determine how to query the underlying issues related to determining the relevance of web pages and queries are web search. Among query to determine the importance of each keyword How high is the key. Right now is a measure of the importance of generic keywords, the principle behind it is information theory.
The most basic technique Chapter 12 Maps and local search
Limited draw maps and local services to use finite state machines and dynamic programming techniques.
These two technologies is machine intelligence and machine learning tools, their application is very extensive, also includes speech recognition, spelling and grammar correction, Pinyin input method, sequence analysis of biological and industrial control and so on.
Chapter 13 Google AK-47 designer - Amit Norris
In all light weapons is the most famous submachine gun. Since it never stuck, easily damaged. It can be used in whatever environment, good reliability. Lethality large and easy to operate. The product is in accordance with the above-mentioned original
The design.
Classification Chapter 14 of the law of cosines and News
Although the computer can not read the news, but news can accurately classify.
The mathematical tools are seemingly unrelated law of cosines.
Two classification Chapter 15 Matrix Operations and text processing in either clustering or classification of text words, mystical value decomposition of the matrix can be carried out by linear algebra. As a result, the problem of natural language processing becomes a math problem.
Chapter 16 fingerprint information and its application
Things in the world has a unique identification feature. Information is also true.
Each piece of information has its specific fingerprint, fingerprints can this difference by different pieces of information.
Chapter 17 of the TV series "plot" have in mind - to talk about the principle
Root password majored in mathematics and information theory.
No information theory to guide the password is very easy to break. There is only after the password information theory are widely used in science, password really become safe.
Chapter 18 flash is not gold - talk about the search engine title
Flash is not gold, search engine ranking pages may not be practical web.
The same principle to eliminate these pages cheating principles and communication filtered noise. This shows that a lot of the principles of information processing and communications are interlinked.
Chapter 19 talk about the importance of the mathematical model
Correct mathematical models in science and vital project, and find ways to correct model is often tortuous.
Correct model in the form generally is simple.
Chapter 20 Do not put all your eggs in one basket - to talk about the type
Maximum Entropy model is a perfect mathematical model.
It can be a variety of information into a unified model. There are anti-cheating wide range of applications in information processing and machine learning. It is very easy in form, beautiful, and in the need to implement mathematical foundation and great skill intensive.
Mathematical Principles Chapter 21 Pinyin input method
Chinese character input process itself is communication between humans and computers.
好的输入法会自觉或不自觉地遵循通信的数学模型。当然要做出最有效的输入法，应当自觉使用信息论做指导。
第22章 自然语言处理的教父马库斯和他的们
将自然语言处理从基于规则的研究方法转到基于统计的研究方法上，宾夕法尼亚大学的教授米奇·马库斯功不可没。他创立了今天在学术界广泛使用的语料库，同一时候培养了一大批精英人物。
第23章 布隆过滤器
日常生活中，经常要推断一个元素是否在一个集合中。布隆过滤器是计算机project中解决问题最好的数学工具。
第24章 马尔可夫链的扩展— 贝叶斯网络
贝叶斯网络是一个加权的有向图，是马尔可夫链的扩展。而从认识论的层面看：贝叶斯网络克服了马尔可夫链那种机械的线性约束，它能够把不论什么有关联的事件统一到它的框架以下。它在生物统计、图像处理、决策支持系统和博弈论中都有广泛的使用。
第25章 条件随机场和句法分析
条件随机场是计算联合概率分布的有效模型，而句法分析似乎是英文课上英语老师教的东西，这两者有什么联系呢？
第26章 维特比和他的维特比算法
维特比算法是现代数字通信中使用最频繁的算法，同一时候也是非常多自然语言处理的解码算法。能够毫不夸张地讲，维特比是对我们今天生活的影响力最大的科学家之中的一个。由于现在基于的移动通信标准主要就是他创办的高通公司制定的。
第27章 再谈文本自己主动分类问题— 期望最大化算法
仅仅要有一些训练数据，再定义一个最大化函数，採用算法。利用计算机经过若干次迭代，就能够得到所须要的模型。这实在是太美妙了，这也许是我们的造物主刻意安排的。所以我把它称作上帝的算法。
第28章 逻辑回归和搜索广告
逻辑回归模型是一种将影响概率的不同因素结合在一起的指数模型。它不仅在搜索广告中起着关键的数据，并且被广泛应用于信息处理和生物统计中。
第29章 各个击破算法和Google 云计算的基础
Google颇为神秘的云计算中最重要的MapReduce工具，其原理就是计算机算法中经常使用的“各个击破”算法。它的原理原来这么简单— 将复杂的大问题分解成非常多小问题分别求解，然后再把小问题的解合并成原始问题的解。由此可见，在生活中大量用到的、真正实用的方法
经常都是简单朴实的。