Slashdot Mirror
Psychologist Beating Math Nerds in Race to Netflix Prize
s1d writes "An almost-anonymous British psychologist named Gavin Potter has suddenly risen to the top of the Netflix prize charts. With his very first attempt, he got a score which took the BellKor team seven months to reach. Currently at a score of 8.07, he has only five teams ahead of him now in the race for the ultimate Netflix algorithm. 'Potter says his anonymity is mostly accidental. He started that way and didn't come out into the open until after Wired found him. "I guess I didn't think it was worth putting up a link until I had got somewhere," he says, adding that he'd been seriously posting under the name of his venture capital and consulting firm, Mathematical Capital, for two months before launching "Just a guy." When he started competing, he posted to his blog: "Decided to take the Netflix Prize seriously. Looks kind of fun. Not sure where I will get to as I am not an academic or a mathematician. However, being an unemployed psychologist I do have a bit of time."'"
5th place is not winning. Are you guys Ron Paul supporters too?
It's called domain knowledge people. It helps being a psychologist when you're write a program reacting to people's behavior. If programmers knew how to do that, they would get laid more.
And yes, I have karma to burn. Yes I do.
It is a miracle that curiosity survives formal education. - Einstein
that had an undergraduate degree in psychology, a masters degree in operations research that after being well employed for a number of years -- "In 2006, he left his job at IBM to explore the idea of starting a PhD in machine learning, a field in which he has no formal training. When he read about the Netflix Prize, he decided to give it a shot -- what better way to find out just how serious about the topic he really was?"
The summary makes 2 references to Gavin Potter being a psychologist, but it ignores the part of the article that notes he has a master's degree in operations research. This is very much an OR problem. Still, it is impressive that he has been able to do as well as he has considering his competition. Good luck to him!
The topic is incredibly fascinating. And just a thought, up to 5 centuries ago "scientists" were incredibly versatile people, with mastery over a few fields at a time... A lot of people argue that this was out of necessity, but could the versatility have been important development of multiple renaissances (In Greece, East/West Asia, and Europe)? And could the bottleneck specialization of fields that has occurred in the past three centuries simply be period of transition/stifling new ways of thinking? Could the emergence interdisciplinary experts lead to another 'renaissance' of sorts?
He might be a psychologist, but his venture firm is named Mathematical Captital, after all. His partners appear to have advanced degrees involving mathematics.
512 MB RAM, 20 GB disk, 200 GB transfer, five datacenters. $19.95/month.
I don't understand why Wired insists on playing along with Potter's pretense of being an "unemployed psychologist". He's a PhD candidate in machine learning, has a masters in operations research, is ex-IBM and Pricewaterhouse, runs a VC firm -- he has plenty of quantitative and computational training and experience, probably more than most of the contestants.
What I'm listening to now on Pandora...
My extensive retail experience says people like to shop by the following methods. Ratings, Genre, Alphabet.
So, if I was to setup a movie viewing for them. I'd setup something along the lines of a Genre, rating(R,PG-13,G), Alphabet.
It's kind of a takeoff on my video game organization method that increases sales of video games by 30 percent. I called it ABSRG short for Alphabetize By Section(4 foot section), Rating(M on top T in the middle and E towards the bottom.), Genre(Sports, driving, shoot em up). Please note, this cannot be patented, I already let it go out for more than a year(Started in 1998)and I have the pictures and time notes to prove it.
I've also found this to be true. Lol, I actually knew people in college that did nothing but program computers in their spare time, and took psych because it was easy, wouldn't distract them and gave them more time to do the programming they wanted. They didn't ever expect to practice psych.
Beware of bugs in the above code; I have only proved it correct, not tried it.
Okay, I'm not trying for this prize, but there's one thing about Netflix "recommendations" that bugs me so I'm throwing out this complete freebie of an idea. If it helps someone get a 0.001% improvement to add this ONE little additional check, great.
I am learning Japanese. I have been watching several hundred Japanese-language movies for the past couple years. I don't watch movies in Greek, Spanish, Turkish, Farsi, Italian, Russian, German, or Hebrew. I did watch Amelie four years ago but that doesn't mean I love French movies. Most of my recommendations are for foreign films, but only a small fraction of those recommendations are for Japanese movies.
Apparently, Netflix doesn't have a column in their database saying WHAT language a movie uses principally, it just has a flag saying it is not English. It's the only explanation I can see for not checking for such a strong correlation. I admit, I might not be sharing the experience of the most common movie-renting drone in the bunch, but I doubt I'm the only person who has such a lopsided taste in movies. If the language (or alternate soundtrack languages) ARE known in the database, please see if the renter has a bias for movies in a particular language.
[
Being beaten up is normal for any nerd, but by a Psychologist - that's gotta hurt...
The dataset is about 660MB to download. It unpacks to 2GB of about 18,000 text files.
What a bunch of drivel. Just because their level of knowledge isn't what we have today, doesn't make it any "easier." Do you have any idea at all, or can you even comprehend, the kind of mathematics that were employed back in the day to solve anything? Take a look at the Principia for example. The geometry is insane. I'm a graduate student in Physics and I can't really follow his proofs.
Furthermore, because early scientists did not have as much to build on, that makes it all the more difficult. Where was Faraday to get his inspiration on lines of force? What lead Maxwell in the right direction to unifying light with electromagnetism?
It's great that 3rd graders know about electric circuits. That's the point of scientific progress. That doesn't make the original task trivial in any sense.
In other words, I hate you.
Yes, the story is old now (by Internet standards) and no, he's not actually winning. What he did do was make a great leap in the success of his process by using a better group of knowledge for a base to work with. This is neither new or amazing in anyway. The only reason it makes news is that it's so much common sense that the story is told as if he has had some huge breakthrough.
Even if he only gets to 9.25% I will bet he gets offers to work with AI researchers around the globe. That is, after all, what their stated goal is - more or less. Every programmer knows about the GUI wars, and has read stories about how programmers have trouble writing code or designing web sites that are intuitive for users. If you want to see you code break, put a user on the keyboard and wait a few minutes.
Just about everything that I do with computers shows me something that could be more impressive or intuitive. Can you say 'click START to shutdown' ? Applying psychology and math to a computer problem is a problem that programmers are faced with all the time, and the industry as a whole fails on this repeatedly. A matter of personal interest, hobby robotics, holds a particular problem that seems simple but is not and demonstrates the scope of the problem with this story. Try to build a small robot that can wander around your house and never get stuck behind the couch, or anywhere else. Even cockroaches can accomplish this, but sophisticated robotics cannot.
We've all seen people come from nowhere, solve a problem because they looked at it a different way than everyone else based on their experiences. I think that it is about time that we started doing more of this. The biggest problem that I can see thus far is that people don't act like computers, they seldom repeat anything with precision. Can you say manufacturing robot? Everyone of us has personal tastes, and it's usually only when Hollywood tells us what movies are good that we all fall in line. Sure, some 'blockbusters' fail, but they make money because of the hype. When you remove the hype, it falls apart. Picking out what other people like or might like based on a very small data set is a difficult task. Not everyone likes kids or movies for kids. Not everyone likes hollywood-ized cookie cutter movies. The task is daunting at best.
Apply that thinking to other things, and you can see why some websites work and others do not. Why some software works and why others fail. Should F1 be the help key or F3? Why not CTRL-H? Maybe your preferences for such things differ from mine. What I'm getting at is that predicting what a human will do is not simple. Categorizing movies by story, style, genre etc. is like applying a tag cloud to it and matching the tag hits of one group to your personal tag choices. It kind of works, kind of does not. Either way, it needs to be applied more often. Just today I received a thank you note from the local Honda dealer where I got my seat belt replaced under warranty. I bought the car 15 years ago from the dealer my mother likes, and is two states away from me now. The dealer that send the card is local to me (2 states from my mom) but they sent the card to her, at MY address. Tell me how a human would have done that?
The basic problem is that we humans accrue various bits of information and make decisions based on that. Our thinking process halts when something 'just doesn't make sense' to what we are doing. Computers don't do that... yet. Perhaps this guy is on to something, but then maybe not. A human would not only ask what other people liked this movie, but also "you really liked that piece of crap?"
To put the I in AI is going to take a lot of rethinking. Simply acting like a perfect human won't do it. Oh, you liked that movie? yeah, me too, I love the city where it was filmed. -- get a program to do that? That oddball out-of-left-field thinking is what will make the software very good at predicting what you will or will not like, maybe.
Have you ever tried to figure out what kind of music someone wou
Support NYCountryLawyer RIAA vs People
There are 480,189 customers that rated 17,770 movies. The total number of ratings that you're given is 100,480,507. Each user/movie/rating is accompanied by the date of the rating, as well. You then have to submit predictions for the ratings of 2,817,131 additional user/movie combos (they tell you the user, the movie, and the date, and you need to predict the rating). You submit these predictions to Netflix, and they tell you the root-mean-square error between your predictions and the actual ratings that those users gave those movies.
There a joke that I heard (that's actually pretty much true):
A Physicist goes to a Mathematician for advise on solving a Differential Equation. The Physicist explains this and writes the equation on a Blackboard. The Mathematician stares at the equation for more than half an hour. Finally, he says, "Yes, it has a solution."
Basically, the Maths (even applied) are about details and considering them *very* carefully. With this in mind, is it any surprise that they are somewhat "slow"? Especially when they are starting from scratch within the problem domain?
OK, how do you solve a cubic? Or grind the lenses for a telescope? Or build a water pump? How do you dress a head wound? Or isolate pure gases? Or calibrate a thermometer?
If you know how to do all those things, good for you. My point is that a secondary school education in "science" does not by itself provide one with the same understanding as the great scientists who first among all others figured out how to do each of these things.
Shop as usual. And avoid panic buying.
Could the emergence interdisciplinary experts lead to another 'renaissance' of sorts?
Oh, I don't know. If only we had a art history & statistics dual major to figure it out...
HA! I just wasted some of your bandwidth with a frivolous sig!
Each of the movies and each of the users represented in the test set have some corresponding known ratings. I think the minimum is around 10 known ratings for users and maybe 3 or 4 for movies. And there are examples on the opposite end of the scale... one user has rated nearly every movie in the training set, and most of those ratings were 1 star. If you have any more questions, feel free to check out my blog - my Slashdot profile links to it, and you can find my email there. I'm Dan Tillberg, currently in seventh place in the contest ;). I'd love to help answer any questions you or anybody else might have (or at least help point you to good resources), and I certainly encourage anybody that's interested to try their hand in the contest :).
The team at the University of Toronto, who are using a neural networks approach, are led by Geoffrey Hinton who has a bachelors degree in experimental psychology.
Subjects get so complex and so specialized there HAS to be limits to how many giants a human can climb and still have time enough to become a giant themselves. Abstraction helps to greatly extend this range; the people behind abstractions/simplifications may not be considered giants because they do not produce progress themselves but just facilitate others so they can extend their reach into the unknown.
There does not appear to be that many 'giant' scientific figures anymore despite the exponential scientific growth. Maybe it is just an appearance and there are more; but are there more proportionally to the number scientists?
How many big leaps in knowledge have been made in the old fields like physics for example? If the decline does not exist now, won't it exist at some point??
Democracy Now! - uncensored, anti-establishment news
Am I the only one that finds the phrase 'An almost-anonymous British psychologist named Gavin Potter' faintly ridiculous?
Eh? How does making use of bog standard free market principles constitute fraud?
Netflix have set out exactly what product they want and the price they are willing to pay. If there are people out there willing and able to supply them with the product at that price, why shouldn't they?
And government employees doing research in thier own time is not costing you a dime.
Netflix benefits through a better product offering, thier customers benefit through a better product, and a fair number of contestants benefit (through exposure and experience if nothing else).
Once a proud programmer of Apple II's, he now spends his days and nights in cheap dives fraternizing with exotic dancers