十分想apt-get这个技能
建模过程:
1.针对的对象不同,也就是选择观测的一个随机变量。能直觉的感到,这个过程本质是自由的,可也不是随意的。
2.选定前提假设。
3.选定模型等式。有时候这个模型等式就是一个假设。
4.用一批数据去验证模型。
5.定制产品影响市场。吸引用户过来付费第一步。要做好博客更新。
原则:
1.分清楚三主体数据
2.分清楚利益主体
3.建立优化目标:
目标1:年度财务最大化=客户数量*产品种类*产品售出数量*产品单价
目标2:客户投资收益最大化
目标3:客户投资收益率最大化
4.建立使得各种目标正向循环的经济循环
学习原则:
1.适合关注问题的最高效的方法
2.快速试验得到结论,推翻自己,建立自己,阶段成果
----------------------------------------------------------------------------------------------------------------------------------------------------------
概率图模型Probalistic Graphical Modle Princeples and Techniques知识结构:
1. Represent Part:
1.1 Bayesian Net Work Represention
1.2 Undirected Graphical Models
1.3 Local Probilistic Models
1.4 Template-based Representation
1.5 Gaussian Network Models
2 . Learning Part:
3 . Action and Decision Part:
Before this three part ,the auther introduced some basic probilistic and graphical knowledge.About Graph I have a question are loops and cycles different?Are Paths and Trails different?
So in my opinions it is just one method to analysis problem.We can do some examples use basic method and knowledge in wiki.
-----------------------------------------------------------------------------------------------------------------------------------------------------------
How to analysis or represent a situation use Graphical Modle?
We look at this article-Simple and scalable response prediction for display advertising.
This is a course from stanford.
After read some materials I even think the Problistic point could come from two part ,one part is our experience and at the same time we can get the Porobilistic points from our learning result.Also we can learn it use our own brain or we can learn it by our learning algorithm and finally we learn the result from our brain,and maybe the process of knowing can be crossed by machine and human`s brain.
Bayesian Net Work is really easy and too many problems can be easily expressed by it,include 马尔科夫概率模型等等。