分享：Rutgers一个老师的讲义

http://stat.rutgers.edu/~madigan/COLUMBIA/
有各种有有意思的topics

Topics in Statistics

David Madigan,
Rutgers University
dmadigan@rutgers.edu or 732-932-7494
Here is a rough outline of the course. I'm not really aiming for a coherent sequence - rather, I hope we can discuss a variety of topics that you might not encounter in your regular classes, and that I hope you find interesting. We will meet Mondays and Wednesdays 1pm-3pm in 903 SSW. During the first class session we will discuss course work, projects, etc.

I am using Professor Heyde's office on the 10th floor. Generally I will be there before and after class for an hour or so, but email me in advance if you want to be sure I'll be there.

Thanks for an enjoyable month! I should be able to assign letter grades tomorrow. Have a good summer! D.

June 5th.
An Overview of Data Mining.
We'll start with an overview of the field of data mining and discuss a few specific data mining algorithms that give a flavor of the way data miners think.
June 7th and 12th.
Logistic and Tobit Regression.
Regularized Logistic Regression with Application to Text Categorization
Brief Notes on Kernel Methods

This class will take a close look at a statistical warhorse that ought to get more attention in data mining. Logistic regression is a particular regression model for binary responses that scales well to ultra-high dimensional applications, predicts as well as many of its competitors, and offers a transparency advantage over many machine learning algorithms.
June 14th.
Brief Notes on Confounding.
Causal Inference in Observational Studies.
Post-marketing Surveillance of Drug Safety (June 19th).

Making causal inferences from observational data seems like the quintessential data mining activity! We will explore the "propensity scoring approach" and related ideas. We will also look at an important application area - drug safety - where these ideas are relevant.
June 19th.
An Introduction to Probabilistic Graphical Models

Probabilistic graphical models uses graphs to represent multivariate probabilistic models. This idea is old but the last decade or so has seen extraordinary developments, both in the appications of graphical models and in the associated rather elegant theory.
June 19th.
Locating Users in Wireless Networks
Causal Inference in "Broken" Clinical Trials

This class will explore two interesting applications of probabilistic graphical models.
June 21st.
Some Graphical Models Theory

We'll take a detailed look at some selected important results in the theory of graphical models.
June 26th.
Text Sequence Modeling

The last few years have seen some interesting developments in models for sequences of words. We'll take a look at some of this focusing in particular on hidden Markov models, maximum entropy Markov models, and conditional random fields. Graphical models have played an important role in these developments. Some of the same ideas have work for biological sequences but I don't know much about this.
June 28th.
???

I'd like to talk about the role of dropouts in clinical trials. Or, we might talk about functional decision trees. Lets see how things go.

里面还有一个１５年老的笑话，今天看来依然很

Ronny Kohavi ICML 1998
The Dream of Data Mining(*)

Dear Mr. Jones:
We noticed you’ve not picked up any condoms at your local
supermarket recently. (Your last purchase was 8 weeks ago.)
Further, you have stopped buying feminine hygiene products,
but have sharply increased your frozen pizza and dinners
usage in the same time frame.
It’s clear that Ms. Jody Sanders has dumped you... We
confirmed this with the Post Office database−−yep, she filed a
change of address...
We at XXX International would like to offer you...
When will we be able to do this?

(*) Variant of a post on rec.humor.funny in 1995

3# 小K

lmao~

现在收集data越来越容易，有能力分析data的人不愁没有工作 ^^

里面还有一个１５年老的笑话，今天看来依然很

Ronny Kohavi ICML 1998
The Dream of Data Mining(*)

Dear Mr. Jones:
We noticed you’ve not picked up any condoms at your local
supermarket r ...
小K 发表于 2010-4-11 05:21

哈哈，这个好逗～

data mining 是很有意思，还很有用～谢谢分享哈～

谢谢分享

点进来发现我居然回过这个帖子.....
ps, 那个链接失效了...不知道楼上是怎么看的........

收藏，谢楼主