Netflix Prize
Is anyone working on the Netflix grand prize of 1,000,000? The one who most improves Netflix's way of predicting movie selection wins. If the improvement is 10 percent or more, it's the grand prize.
Details here:
http://netflixprize.com/
No grand prize candidates yet. Best improvement so far is 5.24 percent, up from 4.96 percent yesterday. Surely with all the programmers here, someone must be working on it, no?
I am closely related to one contestant, and rooting for him of course. He is not posted yet. Netflix has a leader board to track the progress of the top teams.
Details here:
http://netflixprize.com/
No grand prize candidates yet. Best improvement so far is 5.24 percent, up from 4.96 percent yesterday. Surely with all the programmers here, someone must be working on it, no?
I am closely related to one contestant, and rooting for him of course. He is not posted yet. Netflix has a leader board to track the progress of the top teams.
Comments
How the heck can they claim 5.24 or 10 percent better accuracy? Movie preferences change based on the mood of the person choosing. Ridiculous.
That is why it is difficult to do, and Neflix is ofering a cool $1,000,000 to someone who can do it. Certainly the prize is not ridiculous. I'd like to have it.
Why not consider joining the fun? The contest runs until 2011 or so. Check it out. When these discussions get dull, you could be working your way toward a big reward.
This week has seen much more progress than most, going from 4.96 percent or less, up to the current lead with 5.48 percent improvement.
I hope to see results posted by Monday. Interested? Look for Viking under "team" on Monday, but it could be a little later.
http://www.netflixprize.com/leaderboard
Test results show the team will make the leader board, but nothing is certain. Best scenario, tests indicate we might get first place -- not counting on it in any way however. Too much can go wrong, and other teams are making more progress now. If Viking had these results a week ago, there would have been a very good chance of first place, temporarily.
One last comment. I'm a netflix member. The prediction algorithm they use currently is crap. A 10% improvement is almost moot. In order to do anything productive, it would require an extensive behavioral profile of the user/customer, as well as a different understanding of film. For example, instead of "this user likes action movies featuring so and so" it's probably much more accurate to build a correlation between the provided user profile and the actual movie signal, which is not really within out technological frontier for the time being. Doing so would require a lot more memory than we currently have the capablity to utilize efficiently.
Just do like Amazon does -- they've found that in doing certain tasks, like preference prediction, basic image analysis (i.e. is this person female), etc are much more effectively handled by farms of actual people who use a software client that recieves "problems" and reports their answers back to the main HQ for sorting. Seriously: faster, uses less electricity, and everything.
One last comment. I'm a netflix member. The prediction algorithm they use currently is crap. A 10% improvement is almost moot. In order to do anything productive, it would require an extensive behavioral profile of the user/customer, as well as a different understanding of film. For example, instead of "this user likes action movies featuring so and so" it's probably much more accurate to build a correlation between the provided user profile and the actual movie signal, which is not really within out technological frontier for the time being. Doing so would require a lot more memory than we currently have the capablity to utilize efficiently.
They obviously aren't looking for only a 10% increase. That is just the criteria they set for participants to be entered into the sweepstakes. As proven with many other situations, the larger your sample, the more accurate the prediction for the complex system prediction/analysis. The reason why it is impossible to accurately predict the weather with 100% certainty is because in order to do so, you would have to define measurements exactly, which according to chaos theory and quantum mechanics, is not possible: to measure to the degree of preciseness for which is necessary to obtain meaningful or accurate results from a computer system attempting to analyze and predict the complex system's output using the said input.
But the more accurately you define variables and constants, the more acuratly the computer system can predict. And using real world evidence, connecting patterns, and learning from theses connections, is the basis for the way our minds work, and the way we must program to get more meaningful and accurate results in predictions of complex systems.
But the more accurately you define variables and constants, the more acuratly the computer system can predict. And using real world evidence, connecting patterns, and learning from theses connections, is the basis for the way our minds work, and the way we must program to get more meaningful and accurate results in predictions of complex systems.
It's not that I condemn Netflix for holding the contest -- I don't care what they do as long as it doesn't make the product less appealing -- but instead I question the way computers do search and tasks of this nature. I'm not sure if sampling the public is really going to be productive, because if someone or some group actually does build a much better search, they're probably not going to do it within the confines of a 1-million dollar prize. And a much better search it what it's going to take to make Netflix's recommendation pages worth looking at.
. . . if someone or some group actually does build a much better search, they're probably not going to do it within the confines of a 1-million dollar prize.
That's good enough for some of us. Although, a $1,000,000 is not what it used to be.
Go Viking
That's good enough for some of us. Although, a $1,000,000 is not what it used to be.
Go Viking
I've got to say though... seeing as block buster has stores for in-store exchange, return, etc -- I don't see why anybody would choose netflix over blockbuster.
I've got to say though... seeing as block buster has stores for in-store exchange, return, etc -- I don't see why anybody would choose netflix over blockbuster.
It's not necessary to join Netflix to participate in the prize. It happens that I am a member too, at $9.99 a month. I can have one movie at home, but can order an unlimited number, which typically means 8 movies a month. No postage, no late fees, and any movie, even new releases.
But what I like most is, no store. I keep a list of about 45 movies on line, at Netflix. When they get one I send one back, they send the next. If I get it in the mail box for pickup on day one, I get an email on day two saying they got it. Later on that day an email says the next movie is on the way. Day three I have the new movie in the morning mail.
The selection on Netflix seems unlimited. No store could ever hold that many titles. Never has Netflix not had the next movie in the queue to send out immediately, even recently released titles.
I never even considered Blockbuster.
It's not necessary to join Netflix to participate in the prize. It happens that I am a member too, at $9.99 a month. I can have one movie at home, but can order an unlimited number, which typically means 8 movies a month. No postage, no late fees, and any movie, even new releases.
But what I like most is, no store. I keep a list of about 45 movies on line, at Netflix. When they get one I send one back, they send the next. If I get it in the mail box for pickup on day one, I get an email on day two saying they got it. Later on that day an email says the next movie is on the way. Day three I have the new movie in the morning mail.
The selection on Netflix seems unlimited. No store could ever hold that many titles. Never has Netflix not had the next movie in the queue to send out immediately, even recently released titles.
I never even considered Blockbuster.
We'll I know you don't have to be a netflix member to participate in the contest, but I wasn't talking about that. I was merely talking about the fact that blockbuster and netflix offer the EXACT same plans (netflix offers a few more non-unlimited plans, but thats irregardless) at the same price. So the obvious advatage to me is that with Blockbuster, if I want to return a movie, I can go into the store to do that and pick up whatever I want for free. This means no dealing with mail, not having to wait any time for my movie if I want it now, and a larger section if my movies are not shipping immediately from the depot (I've heard netflix prioritizes customers who rent more movies last and not all the movies ship quickly).
So the obvious advatage to me is that with Blockbuster, if I want to return a movie, I can go into the store to do that and pick up whatever I want for free. This means no dealing with mail, not having to wait any time for my movie if I want it now, and a larger section if my movies are not shipping immediately from the depot . . .
Okay, I understand you now.