Saturday, April 28, 2012

2 Thoughts

I saw my 8-month old do the familiar pinched-two-finger-zoom-out gesture of iPhone/iPad when she saw an Anish Kapoor sculpture. Now this has an obvious amusement to it (8 month old, iPad, Apple's genius and so on), but if you know the scale of Anish's work, this is vaguely philosophical.

The other day over research, we said, "Let us go bandit on this". 

Monday, April 23, 2012

Turing Celebrations This Side of the Atlantic

Princeton has been the heart for the past 100+ years, from Einstein to Nash, von Neumann, Turing and many others, pumping scientific blood through the world. Turing Centennial Celebration will be held May 10--12 at Princeton with an excellent program. Enjoy!


Sunday, April 22, 2012

Researchers Will Find a Way

Researchers previously at Yahoo! from algorithms to databases and machine learning, are finding new homes in (Google, LinkedIn, Microsoft, ... ) Researchers find a way. 


Beets on Hands

Cooked beets for lunch today,
Hands and shiny MacBook Air are still red.
The perils of mixing cooking with email.

Thursday, April 12, 2012

Online Free Courses

We know the examples of free online courses with 100k+ students each from Stanford in CS, are there similar developments in Engg/Medical School/Business School/Arts or whatever?


A Claim

If you do applied algorithms, at some point you are going to code/advocate/analyze hubs-and-authority iteration.

Some numbers

Some numbers in news:
  • Microsoft paid AOL $1B+ for <900 patents for an average price of $1.3M. I dont know the max, min and median value of a patent in this portfolio. I have been skeptical in the past with researchers' explaining their super algorithms to me how to ``value'' a portfolio of items including patents, I wonder if any of them would have estimated this value.
  • A few months ago, Mitt Romney released details of his tax returns. One number is on taxable income of $21.7M, he paid $3M in taxes. While media has focused on the effective tax rate of 13.8%, but some of the others focus on a different rate. It is estimated that his net assets are like $200M, but some fraction (<1/2) of it generates taxable income. How does he make 20%+ return a year on his assets?

Sunday, April 08, 2012

Mechanism Design and Commuting

Balaji Prabhakar has been successful running large scale incentive programs in real life in Bangalore and Singapore to alter social behavior (in scale more than a carefully curated marketing study group, less than wild experiment online for 1% of website visitors). Pretty amazing he got these programs running! Now he is running an interesting experiment in Stanford involving congestion and parking relief incentives.


A Puzzle

It has been a while. Consider array A[1..n,1..n] such that any rectangle A[i...j,k...l] has at least one element that appears exactly once in the rectangle. What is the smallest number distinct elements A can have?


Some Almost Technical Thoughts

Ask an Engineer to draw a straight line through 3 arbitrary points on plane, and the Engineer asks, how thick can the line be? People always says this like it is a joke on Engineers, but I really think these Engineers are smart!

Then there is the story of The Hare and the Tortoise. The Hare, confident of winning the race against the Tortoise, stops to look around, smell roses, eat carrots, take a nap, and ultimately loses the race to the slow but steady Tortoise. When I hear this story, I think, if I were the slow and steady Tortoise, I too will still stop, smell roses, eat and nap, along the way.

A colleague, in natural flow of conversation, told me, (s)he was thinking of problems with the``career graph''. There are so many graphs, social, societal, vehicular, .. whatever, why not the career graph.

Open Source DB Algorithms

Joe Hellerstein and others have built MADlib, an open source library for statistical and machine learning analysis with SQL on multiple machines. Algorithms researchers explore and analyze specific algorithms for specific problems (``point solutions'') and a way to use them in practice is to grab the data from database and write your C++/Java Code. In my mind and others', it has been a much-needed intellectual exercise: how and how much of such specialized analyses can be pushed into SQL. Joe and others have taken the exercise further and have built the bones for doing it. Joe is a persuasive thinker, and here is his blog. If you have spare programming cycles, contribute!