Deep learning material

Posted in Uncategorized | Leave a comment

Paper Reading: Low-Rank Tensors for Scoring Dependency Structures

Week Oct 6th ~ 11th


Posted in Uncategorized | Leave a comment

Learning List for 2012 Fall

Copied from prior page

Better than Expect

I came to US in August and started my graduate study in NYC. This is a new start in my career. I have to work harder and harder for the new era. 

I publish my to-learn list in this blog to push myself, for my friends and you are able to see and monitor me!

It’s not the final version, too much time cost. I have to fix in 80h totally.

2012 Fall Learning List

  1. To-Learn Lists, “to be efficient” (86hours~101hours / week)
    1. Required Courses in School (14h/week, up to 20h/week)
      1. Computer Networking,
        1. Homework + Reading
        2. pending, 3h(course) + 2h/week
      2. Computer Architecture I,
        1. Homework
        2. pending, 3h(course)+1.5h/week
      3. Programming Languages,
        1. Pending
        2. pending, 3h(course)+1.5h/week
    2. (15h/week)   (Thanks to professors in coursera, especially Prof. Ng)
      1. Probabilistic Graphical Model,
        1. from 9.24~12.9, 11weeks
        2. Video + Homework + Reading, 7h~10h/week
      2. NN4ML, 10.1*
        1. from 10.1~11.5
        2. Video+Homework, 7~10h/week
    3. In seat Courses (6h/week, up to 8h/week)
      1. Machine Learning,
        1. Homework
        2. 3+2h/week
      2. Operating System
        1. 3h/week
    4. Self Study (6h/week, up to 8h/week)
      1. ~Web Search Engine**
      2. Convex Optimization I, Stanford
        1. 4h/week,Slides+Homework
        2. Itunes U
      3. ~Computational Advertising, Stanford
        1. 4h/week,Slides+Homework,after Convex Optimization
      4. ~Statistics Learning Theory, Peter BartlettBerkeley***
        1. 6h/week, Slides + Homework
    5. Research relative (45h/week up to 50h/week)
      1. Research in Lab
        1. Pending
        2. need to be statistics, at least 20+h/week, almost 30h/week
      2. With other PhD
        1. Interesting Topics!
          1. Detecting in Social Network
        2. 8h / week ***
      3. ACM-ICPC to 11月
        1. 7h/week
          1. will soon over
      4. Review and Study of Algorithm, Book of  “Algorithm” by Sanjoy Dasgupta
        1. 4h/week
  2. To-do Lists, supports a efficient work pace
    1. nice Sleep, 00:00 go to bed every day, at least 1:00 and 50~55h per week
    2. have a noon-snap if possible
    3. 6h sports or gym per week, one time per two days
    4. 5h social event time
    5. eat less and more vegetables and fruit….since food in USA are so much lol

* “~” means I don’t decide whether go on learning, because of time…

** I don’t decide whether to learn this course due to research rate; 2. I would probably take this course next semester

*** I think follow other PhD’s work would help me to learn how to research and get papers.

****maybe this winter, to learn

Posted in Uncategorized | Leave a comment

Using LaTeXit and Tikz-Qtree

I learnt this this morning.

LaTeXit is already installed in Mac. Image

The generated ps could be draged to Keynote


\Tree [.S This [.VP [.V is ] NP ] ]

There should be space before ‘]’s in qtree.

Posted in Tricks | Leave a comment

Online Judges for Python and Haskell Practice

Days before, Professor Huang told me that he was going to open two new courses about Python (Python Programing) and Haskell (Programming Language) in Queens College and Graduate Center of City University of New York. Liang was a genius in Algorithm Coding Competitions (I think every students who attended ACM-ICPC, OI or Topcoder in China know him since he is one of the authors of the famous ‘Black-Book’) I was also a ACMer. During the competitions, we were  both gained much from practice coding on Online Judges. So he wanted to know which OJs support Python and Haskell, so that he coud ask his students to train themselves on them.

I list some OJs below that I know that support these two languages. If you know others, tell me, please 😀

Timus Online Judge: supports C, C++, C#, Pascal, Go, Python, Ruby, Haskell, Scala; This is one of the best OJs; from Russian; Old; very hard.

Zejiang University Online Judge: supports C/C++, Python and Scheme.

Codeforces: supports many languages, traditional ones, and Haskell, Python, Ocaml, Scala, Go; a combination of OJ and Topcoder; from easy problems to very hard ones.

Topcoder: a famous competitive algorithm platform, they hold SRM (single round match) almost every week; supports C++, Java, Python.

HackerRank: supports traditional languages, and Erlang, Groovy, JS, Haskell, Lua, PHP, Python, Scala; a famous Indian OJ, mainly for job seekers.

There are also two new kind of OJs, Project Euler and Rosalind. Each of their problems provides a input file. You could write your program on your computer by any languages, run on the input file and then submit an output file. What’s more interesting, problems in Project Euler are all math problems while problems from Rosalind are all from BioInfo domain. You could learn new knowledge from other domains while doing coding practice.

Last but not least, Python Challenge is an awesome place to train tricky skills in Python.

Posted in Online Judges, Programming Languages | 2 Comments

What are the most important research papers which all NLP students should definitely read?

Which NLP research papers (no matter whether classic or state-of-the-art) do you recommend students to read, if you were a teacher/professor/adviser? And why?

View Question on Quora

Posted in Research | Leave a comment

Slides for WAND, Dewey and Interval Evaluation

I am now an intern @Yahoo! Global. I gave a presentation to my group mates to introduce three algorithms.

Though these Algorithms are core in a Yahoo! Ad and Search engine, they have been published by two papers.

Thus, I cut down the sensitive parts (showed with X in slides) and public this slide (just leave your mail, I will send a copy).

This slide could be used for a newbie guide to get into these algorithms. For deeper knowledge, you have to read the orient papers.

If anyone have any problems, welcome to discuss.

If anyone thinks this is illegal, notice me pleasure.

Leave words and I will give you my email address.

Thank you!


Tech Talk


§Efficient query evaluation using a two-level retrieval process, A. Broder, D. Carmel, M. Herscovici, A. Soffer, J. Zien, CIKM 2003 (pdf)
§Efficiently Evaluating Complex Boolean Expressions, M. F. Fontoura, S. Sadanandan, J. Shanmugasundaram, S. Vassilvitski, E. Vee, S. Venkatesan, J. Zien, SIGMOD 2010 (pdf)
(Just leave your mail, I will send a copy)
Posted in Information Retrieval | Leave a comment