Sunday, June 29, 2008

YouTube Annotations Goes ...

I may alienate my friends going on about youtube annotaions. But here it is. The feature goes human, funny, creative, and increasingly rated-R (ones on the left are interactive, the ones on the right use annotations for adding commentary to the video):

My Bursty Blogs: An Analysis



An anonymous commenter said: "Your blog posts come in bursts. Do you blog when you think of something to write, or do you think of something to write when you blog? For the non-CS posts....Probably a bit of both, like when we write papers...". While I am trying to fill in their thoughts on my non-CS posts, let me respond. I have things to blog about, and jot them down in a notebook like the one on the left or in others blogged about earlier. When I get time (weekends, during travels), I pick one that has the right mass of thoughts and references, and spend 30 min -- an hour writing each of the little posts you see, republishing them a few times. So, I am an inefficient blogger, an old fashioned writer.




I learned from Corinna Cortes: why use informal arguments for analyses when you can use data? Here is the data. Irina Rozenbaum (now at Google) and Hongi Xue (now at Yahoo!) working off a discussion with Graham Cormode built a blog parser that decomposes a blog into posts, comments, words and links and shows you an analysis, such as a plot of number of posts vs day of the week (on the right), list of commenters, words used, etc. The interface is not ready for prime time, so ignore the data quality problems. (The link above will not support the hovering feature.) If you want to get your blogs analyzed, go here. Disclaimer: It has to be a blogger site, and it will take a few minutes. Enjoy!

Johnsons Party

David Johnson throws an annual summer party for the Bell Labs diaspora from prior to '84 and holds together its continuing legacy as a Hydra (the AT&T Labs incarnation has its alumni here). The invite now goes to dozens of organizations, and those in the NJ neighborhood manage to make it to the party. I did, this year, after a hiatus, and was charmed as usual. No photos, just people. Burgers and beers, backyard games. Lot more of interns than I remembered from the past, and the youngsters from a decade ago when I first became a part of this story were there now with their children. Those of the diaspora elsewhere who couldn't make it should know the party might feel familiar, but still it is a subtly varying affirmation to be experienced in person. Here is to the summer!

Saturday, June 28, 2008

WAA08

Even when I worked in a focused/niche area like string or streaming algorithms, it was difficult to find a workshop where I was interested in nearly every presentation like I am now in the Workshop on Ad Auctions organized by Susan Athey, Rica Gonen, Jason Hartline, Aranyak Mehta, David Pennock, Siva Viswanathan. The workshop is nestled within EC, AAAI and World Congress of the Game Theory Society, amidst the fantastic architecture of Chicago, and the landing of the incredible artistic hulk of Jeff Koons at the Museum of Contemporary Art.

Sunday, June 22, 2008

Wines, Cameras, Car Dealer, Etc

Some of the researchers may be focusing on the WINE conference deadline, but YouTube Annotations is helping people:
  • Sommelier explain decanting with annotations in Italian. Enjoy Aldo Sohm.
  • Canon HF10 in Japanese with English (and ROMAJI: romanization of Japanese) translation using annotations. Cool.
  • A rapid catch of the police. Chop chop.
  • Finally, car dealers too get into the game selling an Acura (watch in high quality); the annotations are really detailed. Nice.
ps: Cheers to Youtube Screening Room. Go, indie filmmakers.

Mizoguchi Treat



Last Saturday I watched this movie at the IFC, a rare treat for me at 11AM: a moment from 1954, black and white, still and nearly silent, haunts no matter how many times I watch it. And for the next four weeks, Film Forum has taken away any reason to work with its retrospective of Tatsuya Nakadai's work.

Labels:

AGT Primer

I got to hear Tim Roughgarden speak about recent results, in particular, his recent STOC and SAGT papers. He gave a nice technical description of Meyerson's optimal auctions result and argued how the probabilistic utility model there is no inherent limitation as his results show for classical TCS type prior-free optimal mechanisms. Fantastic stuff, and Tim is a powerful speaker. He has a STOC/FOCS type primer on Algorithmic Game Theory that is now online. Enjoy!

Saturday, June 21, 2008

Finding a Way

All blogs are self-indulgent. Apologies: this is a beyond self-indulgence, totally non-CS, non-whatever.

I celebrated the end of the week by subwaying down to Brighton Beach on friday for a Russian evening where I watched Tsygane (one of the many Russian gypsy, folk groups operating in NY) at the National club. I was distracted thinking about the gypsies and and other marginalized groups.

I am sure many of us can recall the quote "Life will find a way"; I tend to think in shorter time scales of generations, lifetimes of people: "People will find a way." Societies have pushed various ethnic/racial/social/sexual groups into corners, encased their stereotypes into "jokes", pushed them aside, or prosecuted them using some pseudologic or the other (that still persists) ; these groups have found themselves taking to the stage, performing, distorting themselves sometimes grotesquely, and ultimately they coexist with their prosecutors in this cloak of "entertainment".

Gypsies left India in the 11th century, gathered together the food, music, and the languages of Middle East, Europe and Northern Africa during their epic migration over the past 900 years and wear it like the patchwork of colors and sequins on their clothes; they are yet to find an unpromised land or time, or a debate that resolves, or a society that will pass the test.

Labels:

Tuesday, June 17, 2008

No Moss

Corinna Cortes, who has done a lot to put together the research group at Google NY, verbalizes a notion: the need to take responsibility for a problem, project, whatever. Researchers, in general, find that challenging. We feel the ultimate responsibility for our result, which may be a theorem, an idea, a concept or a system design. But beyond that, it is a challenge to feel the responsibility for the underlying task (in say vision, speech, databases, networking, economics, ..). We'd rather seek the next extension of our result (new theorem, different approach, better design). We are rollin' stones.

On Being a Tomato

Applied Algorithms Researchers are caught between on one hand their drive to be problem-solvers and theorem-provers, and the other hand, the need to peel away the stated motivation and peer behind it to become familiar with the application. Danger is, they may become a tomato which is neither a fruit nor a vegetable.

ps: I really liked a paper on String Folding, by Mike Paterson and Teresa Przytycka, where the authors show great integrity by resisting the temptation to call it "Protein Folding". Do this mental exercise when you read a paper: replace the affective name for the problem ("finding network anomalies" or "signature" or "suspicious people") by its effective name ("the IP address that sent most traffic" or "most frequently visited websites" or "clique in a communication graph"), and see what ensues.

pps: To those who care, the name Las Ketchup is derived from their father, who is known as El Tomate.

Van Gogh Medicine




When I was sick with what looked like flu, friends (you know who you are) adviced me to drink this or that to get better: vodka straight, vodka with lemon and honey, vodka with pepper, vodka with horseradish, vodka with spicy... etc. Either vodka is the most therapeutic drink out there or my friends are vodka-holics. :)

Labels:

Economic Design Conference

I am at the Society of Economic Design conference. Mohammad organized an invited session on Sponsored Search which included talks by Robert Kleinberg (he spoke about on his result with Aaron Archer on characterization of truthful auctions, the result looked really intricate) and Michael Wellman (who spoke about empirical game-theoretic analysis that seems quite relevant to some of the things I think about); I, an outsider, was happy to get backdoor access to this conf of Economists. I discovered there is a lot in common between CS and Econ theorists. They too have "reductions": these reductions put some structure on mechanisms so you can reason about whole class of them using a generic one. I met several economists like Utku Unver who seemed really open to interaction with algorithms researchers. There was an hour long talk by Susan Athey who is moving to spend some time at Microsoft Research; her talk was tutorial-esque and she was somewhat zealous in getting the M/S message out. Still it is good to see Economists getting interested in internet ad auctions research (After my talk, someone said, "Wow, I did not know there was so much thinking behind that text on the right side of a search page!"). I was curious to see if Economists had any offbeat new topics for research. A candidate seems to be neuroeconomics: the study of neural activity around economic decisions taken by people.

Tuesday, June 10, 2008

Build It, Release It on the Web and They Will Come

There are moments that one feels like some mission has been accomplished. Youtube annotations was a lot of work, it basically makes youtube videos hyperlinkable via popup annotations (the web was basically text made hyperlinkable, and then it became more). And now in a satisfying development, people have used it to link together several Godfather videos. This many-layered movie (I, of course, II is mainly a rear view mirror, and III was the collapse) continues to occupy people's psyche.

People also use the feature to do fireworks, get help choosing clothes, or even attempted) spam.

Monday, June 09, 2008

Food discoveries

During the weekend, I discovered Jachnun (yemenite breakfast), and in following that lead, found this wonderful breakfast blog that is searching for the best eggs in town. Enjoy!

And sometime ago, I discovered Molokhia: a Jute soup from Egypt. Jute is a theme from the past for me. Jute connotes the partition of the larger Bengal in'47 that left Jute fields on one side and the processing factories on the other, and created a stasis that has held since.

Labels:

Saturday, June 07, 2008

Sponsored Search Tutorial

Jon Feldman and I gave our SIGMETRICS tutorial a few days ago. Sponsored search is often thought of as a game between two parties: the advertisers and the search company. But implicit in this is the third party: the search users. I did the overview at the beginning of the tutorial laying out this three party game explicitly; Jon presented technical results on the advertisers' and search engines' points of view and I presented technical results on the users' point of view. The room was small, but the audience spilled over (we had to go up against Ed Coffman's keynote in parallel). The audience was divided between the ones who wanted an overview to the sponsored search ad system and those who wanted to get to the submdular optimization soonest.

We ran an ad related to our tutorial with Google and showed the statistics of clicks, impressions, click-thru-rate etc. The details are enclosed.

Rejuvenate

I have been down the past week, sick some, worn down some from work, and haunted some by things not done and the chores that remain on the balance sheet.

This morning I reached for the New Yorker, too tired to hope. I liked the calming print on an article. A solitary runner with purple hills in the background, creeping greenery on the side and orange shoes in their stride. I started reading it and something about the writing tickled me. My nostrils felt it first. Like a man perking up to the fragrance of coffee in the morning, the rest of my senses followed. My eyes opened. I nudged my body to a more comfortable position on the low couch, and finished reading. Life and Letters, by Haruki Murakami. In this wonderful article, the writer discovers that he is a novelist. And I who long ago identified him to be the novelist, discovered he was a long distance runner.

With a few simple sentences, the author stirs up memories some, offers advice some, and slips in words into my psyche some as I start on the weekend:
At any rate, this is how I started running. Thirty-three---that's how old I was then. Still young enough, though no longer a young man. The age that Jesus Christ died. The age that F. Scott Fitzgerald started to go downhill. It's an age that may be a kind of crossroads in life. It was the age when I began my life as a runner, ...

Wednesday, June 04, 2008

Annotate that!

YouTube now lets you watch video that has been annotated, click on annotations and interact with the video. The feature went live yesterday, and the blog world has a number of articles 1, 2, 3, ... and examples 1, 2, 3, ... including a how-to (how-to video). Now video uploaders can annotate the videos, set up branching storylines on top of their video collection etc. Let the fun begin!

ps: Loved this test drive. Watch it on high quality.

UPDATE: If you build it, people will play, hack, subtext, and lyric Britney (I am waiting for a Shakira).

Sunday, June 01, 2008

Recycling

The math was George Lucas + Steven Spielberg + Harrison Ford: how can they go wrong? I went to see the new Indiana Jones film and discovered that math really equals (old man-young apprentice, dad-son revelations and sword fights) plus (alien skull, close encounters of the third kind, and ETs go home). Special effects of the worst kind, definitely TV-quality. I walked back via the ever-changing mess of St. Marks place:

St Marks is a mess,
And was always a mess.
But now the theory goes
That lower east side is a mess,
And Williamsburg is, and so's
Brighton Beach, I suppose.
The dear only knows
What will next prove a mess.
St Marks, of course, is a mess
But was always an exciting mess.

ps: I just an ad for a Karaoke bar called, uh, Sing Sing, which to locals, means this.

Labels:

Now for something completely different

Some of the readers may be interested in databases research, or at least in algorithms in databases research. I just updated my algo.research page with the link to a paper on Database Research Projects at Google. Enjoy!