Netflix Prize: The Competition

Loading...

Из курса от партнера Princeton University

Networks Illustrated: Principles without Calculus

40 оценки

Princeton University

Networks Illustrated: Principles without Calculus

40 оценки

What makes WiFi faster at home than at a coffee shop? How does Google order its search results from the trillions of webpages on the Internet? Why does Verizon charge $15 for every GB of data we use? Is it really true that we are connected in six social steps or less? These are just a few of the many intriguing questions we can ask about the social and technical networks that form integral parts of our daily lives. This course is about exploring the answers, using a language that anyone can understand. We will focus on fundamental principles like “sharing is hard”, “crowds are wise”, and “network of networks” that have guided the design and sustainability of today’s networks, and summarize the theories behind everything from the social connections we make on platforms like Facebook to the technology upon which these websites run. Unlike other networking courses, the mathematics included here are no more complicated than adding and multiplying numbers. While mathematical details are necessary to fully specify the algorithms and systems we investigate, they are not required to understand the main ideas. We use illustrations, analogies, and anecdotes about networks as pedagogical tools in lieu of detailed equations. Please note that per Princeton University policy, no certificates, credentials or reports are awarded in connection with this course.

Из урока

Movie Recommendation on Netflix

One of the perks of having a Netflix subscription is getting recommendations of movies to watch. Behind the scenes, Netflix uses powerful algorithms to determine which will be suggested to each person specifically. In this lesson, we will take a look at the main ideas behind these algorithms.

Netflix Timeline5:21

Video Streaming5:19

Recommendation is Everywhere3:24

Netflix Recommendation System5:20

Netflix Prize: Logistics7:38

Netflix Prize: The Competition4:33

Our Example3:48

Raw Average7:13

User-movie Interactions7:00

Baseline Predictor6:57

Cosine Similarity9:18

Similarity Values10:58

Leveraging Similarity8:30

Neighborhood Predictor4:44

Performance of Different Methods3:30

Copy of Cosine Similarity9:18

Познакомьтесь с преподавателями

Christopher Brinton
Lecturer
Electrical Engineering
Mung Chiang
Professor
Electrical Engineering

So now we can take a look at the competition itself.

This is just a timeline from 2006 until ending around June/July 2008.

So it started again in October 2006 and in fact within a week of the start of

competition CineMatch is beaten by 1%. So one team had actually gotten 1.06%

improvement. So in this graph right here we , the

actual values that are plotted the 0.9514 that was CineMatch's original value.

This 0.9413 is where they were a week after.

And that's a 1.06% improvement over 0.9514.

So we're referencing this all to 0.9514, so this one's 8.26% better than this.

This one is 8.43% better than this, just to give an idea.

Now, at the end of each year, Netflix had what were called Progress Prizes, and it

would, it would award a certain amount of money.

I think it was $50,000 to whichever team had gotten the, the most improvement over

that year, basically. And there were only.

O-, one of the only, the only requirement for that was it had to be at least 1%

over the previous year. So in order for you to qualify for the

Progress Prize your algorithm had to beat whatever was in the Progress Prize the

previous year by at least 1% and it didn't it only went on for you know two

years. But so the 2007 Progress Prize was won by

this team called BellKor and BellKor had reached a 8.43% improvement.

Right. So, you see already.

By 2007, they had already reached 8.43% improvement in the RMSE and this was

about a week before they, final call for the Progress Prize.

So, what happened was they would say okay, well, last call you have 30 days to

make your final submissions and. And then at the end of that then they

would run and see who was doing the best. And at that point, BellKor had had an

8.43% improvement. And that was the best one so they won

that Progress Prize. So you can see we're already over 8%

improved over CineMatch. So setting 10% as the target goal.

They didn't necessarily notice, they probably had a really good idea.

But 10% was, a really good target and you can see clearly because if they hadn't

done that then within one year, if it was like 8% or 5% or something, someone would

have already beaten it. But it still took an extra year on top of

that before they actually could beat and get the amount of improvement.

So in 2008 the Progress Prize was won by the team called BellKor in BigChaos, and

so it was actually BellKor which merged with another team called BigChaos.

And then they form this super team, so to speak, and they got 9.44% improvement.

So, one thing, again, that you'll notice is that we jumped from 1% to 8.43%, from

2006 to 2007 then 2007 to 2008 we only improved by like another percent.

So you see that the last few percentages are really hard to squeeze out in that.

And this actually determined that 2009 was going to be the last year, because

remember in every year you had to improve it by at least 1% over the previous year,

so they needed to get at least 10% by the next year.

So now the, the grand finale and this is pretty interesting and funny story

actually. There's this other team called the

Ensemble, which reached 10.10 on the quiz set, so they.

BellKor, it was actually by this time called BellKor's Pragmatic Chaos, because

they merged with another team called Pragmatic Theory.

They had gotten 10.05%. then they enter the last call phase.

Then the ensemble had gotten 10.10% on the quiz set by the end of the last call

phase, so it's just slightly better, right.

And, then they entered the, the final call, where you just had to just submit

your algorithms and, it turns out that the ensemble and BellKor's Pragmatic

Chaos both got 10.06%, 'kay. But BellKor's Pragmatic Chaos submitted

their algorithm 20 minutes earlier. So this is again, this is on the quiz set

and in fact the Ensemble had a slightly better performance on the quiz set than

BellKor's Pragmatic Chaos did. But they both tied at 10.06% on the test

set. But BellKor's Pragmatic Chaos submitted

the algorithm 20 minutes earlier, as so they were the winners of the million

dollar prize and the Ensemble lost for that reason and so a very exciting 24

month period came down to a 20 minute differential.

Ознакомьтесь с нашим каталогом

Присоединяйтесь бесплатно и получайте персонализированные рекомендации, обновления и предложения.

Coursera делает лучшее в мире образование доступным каждому, предлагая онлайн-курсы от ведущих университетов и организаций.

© Coursera Inc., 2018 Все права защищены.

Загрузить из App Store

Загрузить в Google Play