Print This Post Print This Post 1,463 views
Mar 02

Just visit Kaggle is a platform for data prediction competitions that allows organizations to post their data and have it scrutinized by the world’s best data scientists. See how it works.

Who should host a competition on Kaggle?

Crowdsourcing data modeling is an effective way to build predictive algorithms. There are any number of approaches that can be applied to a data modeling problem, but it is impossible to know at the outset which will be most effective. A consultant or in-house statisitician can try a few, but opening up the problem to a wider audience ensures that organizations reach the frontier of what is possible from a given dataset.

Most data problems can be framed as a competition:

  • > banks predict which loan applicants are likely to default
  • > government treasuries forecast tax revenues
  • > websites make personalized product recommendations
  • > hedge funds crunch data to find trading opportunities
  • > bioinformaticians search for links between genetic markers and disease

Who should compete on Kaggle?

Data scientists rarely have access to real-world data. This is frustrating when you consider that many of the world’s organizations have piles of data that they can’t make the most of. Kaggle corrects this mismatch by giving data scientists access to real-world data and problems. Best of all, the burden of collecting, cleaning and structuring the data will have been done by others.

Competitions offer participants the opportunity to:

  • > try their techniques on real-world problems and receive real-time feedback
  • > enhance their reputations and win prizes
  • > meet and collaborate with new contacts
  • > improve their skills and help to rapidly progress the state of the art in a number of different fields
Tagged with:
Print This Post Print This Post 1,155 views
Jan 18

Reported by By Bob Brown, in Network World, January 11, 2011.

Google Science Fair looks to bring glory to science talent.

Google (NSDQ: GOOG) is urging youths ages 13 to 18 to take part in a worldwide science fair that will be hosted by the search giant online.

Google, in a blog posting titled “Google Science Fair seeks budding Einsteins and Curies,” invokes the story of its founders, onetime computer science students Larry Page and Sergey Brin, to encourage young people to take part in the event.

“Larry and Sergey were fortunate to be able to get their idea in front of lots of people. But how many ideas are lost because people don’t have the right forum for their talents to be discovered? We believe that science can change the world—and one way to encourage that is to celebrate and champion young scientific talent as we do athletes and pop idols,” Google writes.

The Google Science Fair is being conducted in partnership with CERN, The LEGO Group, National Geographic and Scientific American.  Details on how to enter are here, but the basics are that students can enter by themselves or in groups of three by April 4.  Finalists will be invited to participate in a live event at Google headquarters in Silicon Valley. Prizes include everything from a trip to the Galapagos Islands to scholarships, and entrants are free to double dip by submitting projects they are doing for local competitions into the Google Science Fair.

The Google Science Fair isn’t the first time Google has sought to inspire creativity via the contest route. It used to hold an Android Developer Challenge to entice programmers to create apps for Android smartphones. (That effort seems to have worked out pretty well, given the growing popularity of Android devices.)

Tagged with:
Print This Post Print This Post 1,305 views
Jan 12

Reported by NewScientist, 27 December 2010 in 2011 preview.

A draft solution to the so-called “P versus NP” problem generated excitement in 2010 – will 2011 bring a correct proof?

Vinay Deolalikar made waves in August when his draft solution to a mathematical problem that haunts computer science hit the internet.

It’s known as “P versus NP”, and a correct solution is worth $1 million. Sadly for Deolalikar, of Hewlett-Packard Labs in Palo Alto, California, his work didn’t check out. But the flurry of online activity surrounding the paper demonstrated a new way of doing mathematics – via blogs and wikis – and generated fresh excitement around the problem.

Formulated in 1971, P versus NP deals with the relationship between two classes of problems that are encountered by computers. P problems are relatively easy for computers to solve. But it can take an impracticably long time to solve NP problems, such as finding the shortest route between several cities – though it is easy to show whether a possible solution is correct.

If P = NP, computers may eventually be able to solve a host of complex problems, from protein folding to factorising very large numbers. The ability to solve the latter would spell trouble for algorithms that we rely on for internet security. Most people assume the opposite is true, that P ≠ NP; had Deolalikar’s paper been correct, it would have proved this.

The Clay Mathematics Institute in Cambridge, Massachusetts, has promised $1 million to the first person who can prove it one way or the other. To find out how likely this is to happen in 2011, see “Prediction: Its time has not come”.

Prediction: Its time has not come

Unlike many problems in science, highly theoretical enigmas like P versus NP are rarely solved piecemeal. Instead, they tend to remain unsolved for years and then, apparently out of nowhere, a proof that works pops up.

Predicting these breakthroughs might seem impossible, but we devised a way to estimate the likelihood of P versus NP being solved next year. We compared its “age”, or the time since the problem was formulated, to other long-standing mathematical problems.

First we compared P versus NP with 18 mathematical problems, from Fermat’s last theorem to the Poincaré conjecture, that were not solved until more than a decade after their “births” (see graph). This made arriving at a solution to P versus NP in 2011, when it will turn a sprightly 40, look premature: just 22 per cent of these other problems were solved before they turned 40. By the same logic, in 2024, we should be on the lookout for a solution to P versus NP. That’s when it turns 53, the age by which 50 per cent of the problems we examined were solved.

Here’s hoping that solving P versus NP turns out to be faster than proving the Honeycomb conjecture, which states that if you need to divide a surface into tiled shapes of equal size, a hexagon is the shape that requires the smallest length of dividing lines. Proving that took more than 1500 years.

We also compared P versus NP to 26 other problems that still haven’t been solved. In 2011, it will be younger than 81 per cent of those. Samuel Arbesman and Rachel Courtland

Read more: “In with the New Scientist: Our predictions for 2011″

Tagged with:
Print This Post Print This Post 1,233 views
Nov 20

IEEE Computer Society is presenting the 2011 Simulator Design competition for students worldwide with a top prize of 8,000 USD and a second place prize of 2,000 USD. Student teams will be invited to design a CPU simulator, a program used in many architecture courses to illustrate how computers work.

“This is an exciting competition because it cuts across traditional boundaries by combining architecture with program design and software engineering – just like real life,” said Alan Clements, chair of the competition and an emeritus professor of computer science. “All you have to do is to write a program. Well, that’s not quite all. You have to write an excellent program using professional design techniques.”

The competition requires that students have taken a course in architecture and have both programming and software engineering skills. Student teams will submit both a report and a working program at the end of the competition.

Who can compete?

The competition is open to student members of the IEEE Computer Society organized into teams consisting of three to five students enrolled at the same institution of higher learning.

Current IEEE student members can add Computer Society student membership (8 USD in the U.S. and Canada and 13 USD in the rest of the world) at:
http://www.computer.org/addcsmembership

Non-member students can join both IEEE and IEEE Computer Society (40 USD) at:    http://www.computer.org/studentoffer

As part of their member benefits, all student members receive access to the Computer Society Digital Library (CSDL).

The competition is conducted through online submission of reports and simulators to the panel of international judges (chosen by the IEEE Computer Society).  This year’s judges include Bob Colwell, one of the world’s leading experts on computer design and Intel’s former chief architect on the Pentium 4 processor.

To register and for more information visit the competition web site at:  
http://www.computer.org/portal/web/competition
(Registration deadline is 18 January 2011)

Tagged with:
Print This Post Print This Post 1,345 views
Nov 13

Reported by George Musser Nov 1, 2010 in Scientific American.

The Foundational Questions Institute announced this week its latest essay contest, “Is Reality Digital or Analog?”, and if it’s anything like the past two contests, we’re in for a real treat: the contest should draw entrants from some of the deepest thinkers of our time. This time around, Scientific American has joined the institute as a co-sponsor of the contest.

The article we published in June on the nature of time, written by philosopher of physics Craig Callender, grew out of FQXi’s first essay contest. The contest, which has a first-place prize of $10,000, is one of the ways that FQXi—a smallish, newish organization that gets money from the Templeton Foundation and other donors—supports cutting-edge research that tends to fall between the cracks at larger, risk-averse funding agencies.

An essay contest might sound like something you’d have done in high school, but such competitions have a distinguished history in science. Cash is always welcome, but the main benefit for most participants is the opportunity to play with an idea in a way they can’t in a formal journal paper. The Gravity Research Foundation, for example, has run one since 1949, and practically everyone who’s anyone in gravitational theory, from Steven Hawking to Roger Penrose to John Wheeler, has entered. Scientific American itself ran a famous essay contest in 1921 to explain Einstein’s theories of relativity.

FQXi’s contest on the nature of time and a second one on the limits of physics drew a huge variety of fascinating contributions from a veritable Who’s Who of physics. It also allowed for new voices who might not otherwise get heard. This is one of the few times when an institution of science is willing to run the risk of psychoceramics in order not to exclude potentially interesting ideas. All the entries were posted to the FQXi website and anyone was free to comment on them and vote for a winner, in addition to the selection of a panel of judges.

As every essay-writer knows, half the fun is to interpret the question. The latest, about digital vs. analog reality, could go in a lot of different directions. The obvious one is to ask whether spacetime is discrete and what that would mean, but I imagine that entrants will come up with even more interesting interpretations. In our November 1999 issue, cosmologists Lawrence Krauss and Glenn Starkman posed the digital-vs.-analog question and discussed what it meant for life in the very far future of our universe.

The FQXi scientific directors, Max Tegmark and Anthony Aguirre, and I have been trying to get our two organizations to work together for several years, but it only came together this year. We brainstormed enough essay questions for the next dozen contests, we’ll work together on the judging, and our hope is that the prize-winning essay(s) will appear in some form in the magazine. In the meantime, bookmark the contest site and check back periodically to read what entries people have submitted!

Tagged with:
preload preload preload