Big Data Way Hmm series < two > (growth)

One: Hmm decoding problem

(1) The depth of programming is nothing more than algorithms and structures, as well as various architectures and versions of management (such as Git management), so as a programmer algorithm this level is not around the past;

(2) about the algorithm, the personal comparison advocating a book is "Introduction to the algorithm" and the ACM real-combat system of the algorithm training;

(3) for the Natural language processing field or some of the mechanical learning domain algorithm, hmm model is one of the most classical algorithms, it is very suitable for beginners to learn and research;

(4)hmm model μ= (a,b,π), the state is not visible, we see only the state of the observed values and the state probability function.

Two: Introduction to hmm (U = {N,m,a,b,pai}---> U ={A,b,pai} )

(1) Number of States in the model N

(2) Each state can output the observed value m

(3) State transfer matrix A={aij}, where AIJ satisfies AIJ = P (Qt=sj|qt-1=si), and AIJ sum of J is 1

(4) Emission matrix B = {BJ (k)}, i.e. the probability distribution matrix from the state SJ observation to the symbolic VK

(5) Initial state probability distribution Pai

Three: three basic questions of HMM

(1) Assessment question: Given an observation sequence O=o1o2 ... OT, and Model μ= (a,b,π), how to calculate the observed sequence O=o1o2 ... The probability of OT, i.e. p (o|u)  

---------> Forward backward Algorithm forward-backward algorithm (forward vector and back vector, one formerly counted, one from the back)

(2) Decoding problem: given an observation sequence O=o1o2 ... OT, and Model μ= (a,b,π), how to quickly and effectively select the "optimal" state sequence q=q1q2...qt in a certain sense, so that the state best interprets the observation sequence.

---------> Viterbi Algorithm Viterbi algorithm

Take a look at the big data of the road HMM series < >

(3) Learning problem: given an observation sequence O=o1o2 ... OT, how to tune the parameter μ= (a,b,π), so that P (o| M) Max

---------> forward Backward Algorithm Forward-backward algorithm (forward vector and back vector, one formerly counted, one from the back)

four: Knowledge Supplement (partially reproduced)

(1) as a good programmer, you need to master a text editor/IDE and persevere in mastering the tool 。 If you choose VIM, then become its owner and let it go east to the East

< Span style= "font-family:arial; line-height:26px "> (2) using the version control system, I recommend using Git. Here's my favorite resource: Pro git--'s free online e-book, which explains all aspects of git in detail.

(3) will refactor the habit,

(4) Learning software architecture, Learn about the architecture based on the stack you chose above. I recommend Martin Fowler wrote "Patterns of Enterprise Application Architecture" book

Five: Chicken Soup of the Soul (partially reproduced)

(1) A brief analysis of the------> of the thought of science students and liberal arts students

(2) One-yuan mode of thinking: Beauty is the United States, ugly is ugly, the two have a principled difference, as if judging good people and bad people, there must be a standard. They like to make all kinds of "objective standards", and also believe that there is a best way to solve problems, if they do not know, in others. Moreover, this mode of thinking people like to pursue "to the high" "to the United States" "to the Big" the "realm", will continue to pursue in one direction, until can not continue.
(3) Two Yuan mode of thinking: Beauty and ugliness are relative standing and exist, can be transformed into each other. Just like good people and bad people, they exist relative to each other and are transformed under certain conditions. This person's life, and character will become more harmonious, easy to understand others. Such people are more likely to become managers.
(4) Multi-mode of thinking: The world is not beautiful ugly good and evil, all depends on the people's thinking and judgment angle, this mode of thinking is superb, with very high flexibility; This is where you are in [Lao Tzu]

(5) mathematical logic thinking is flawed: it is only a one-dimensional thinking mode, lack of flexibility. and a meta-mode of thinking easy to go to extremes, the so-called "Dead End", "a way to the black", is such a person; "stand firm, the views are clear." The advantage is focus, minded. So many engineers and technicians seem to be more rigid, life is no fun. Some of China's outstanding retired engineering and technical personnel, amateur life can be said to be very miserable, many people and children's family relations are not good, very lonely. But the best scientists, such as Einstein, do not, because their way of thinking has jumped out of a dollar into the plural world.

only the liberal arts can provide a multi-thinking approach to education. If received a good liberal arts education, most people have the opportunity to learn two yuan way of thinking, is the so-called "in Split", to reach this level of the person will not be extreme.

In short, as a technical background, we all day to the machine's program Ape : suggested to read more articles, to read good articles, rather than the book of romance, read some more logical articles, will be helpful And in the course of reading the story, try to speculate on the contents of the following article, which is very helpful for trying to figure things out. There is to read the article do not rush to understand, a good article is often read the number of sides more understanding of the more understand. Also have to write more articles, also have certain help. That's the way I study literature in general. So the general test answer analysis and debate types of problems are often higher than others. The liberal arts is the accumulation of learning can not be anxious. To develop slowly . Especially in the area of thinking. Look more on the thought of more, think more of the answer is more natural.

Big Data Way Hmm series < two > (growth)