Name: 5 14 潜在因子モデル (5 14 Latent Factor Models 16 11)
Uploaded: 2021-01-14T07:02:43.000Z
Duration: 16 min 12 s
Description: VoiceTubeの動画で発音を聞きながら英語表現を覚えよう！学べる英語：

Anant has already talked to you about recommender systems, and so far we have

to 不定詞

learned how to do, collaborative filtering recommender systems and

命令形

how to do, profile based recommender systems.

we will look into latent factor recommender models.

So basically the idea will be that we will adopt this

machine learning perspective where we'll be using optimization in order to build.

And in particular, our lecture today will be moderated by, by,

So, Netflix is, is a company that is renting out, DVDs, and in order

to increase customer experience, Netflix wants to be able to recommend which

customers will like which movies, and this is called the movie recommender system.

Right, and a few years ago, actually, Netflix prepared some data and

created The Grand Challenge, where the idea was who can build the best

recommender system and the person who does that gets a one million dollars prize.

So the idea was that first we would get, scientists would get training data.

This is a set of data of 100 million ratings from 480,000, users.

So based on this data the idea is that we build.

Or create a recommender system and then we would send these

recommendations to Netflix that would check our predicted recommendation.

Basically how much does a given user like the movie.

We did their test data set which is a set of Ratings that only Netflix has, and

this way we would want to predict or, the, and make sure that the predicted ratings

are correspond to the actual true ratings of the people who like those movies.

So, we are given hundred million movie ratings, and

we want to build a recommender system out of this.

So, the way we can think of our ratings data in,

And, basically, the way this matrix is composed is that

rows corresponds to movies and columns corresponds to users.

most of it is empty in a sense that most users haven't seen most of the movies.

Right? But whenever a mo, a user sees a movie and

rates that movie, that represents an entry in this matrix, right?

So for example, this entry, in this particular en,

element of the matrix, number 5, that would mean that, a user in the,

in the fourth column liked the third movie a lot and rated it given five stars.

The way we think of the recommendation problem is in a sense that we want to

predict how much will a given user like some movie that

So in this sense, we can use the historical rating data to build our model.

This is what we'll call the training data.

And then using this strange model we want to predict how much the users like

movies that they haven't yet seen and we will call this the test data.

And now when I say that we want to predict how much will a user like a given movie

the way we want to quantify how accurate is our prediction we will use what is

So, basically the idea will be that our goal is to make predictions about how

much users are going to like the movies, then we will go and ask these users hey,

And then we want these two, the prediction and

the true value, to be as close as possible.

So, the root mean square error, the way the formula works, is follows.

For a given user x in movie i I want to predict how much will that

user-movie pair what will be the rating of user x to movie i.

Of course I also know the true rating for movie i of user x.

And what we want to do is we want the discrepancy between the predicted rating

To be, to be small in a sense that we want to take the difference

square is because the difference could be either positive of negative.

So if we square it up, everything is positive.

Now we sum all these errors across all the users and all the movies.

We take the square root and divide by the total number of ratings, and

So in some sense, our goal is to in some sense complete this matrix R.

Wherever the values for ratings are unknown.

And our goal is to complete this matrix as accurately as possible.

we want to predict how much is a given user going to like a given movie.

So what is the input and what is the output, right?

So what Netflix gives us is the training data of hun, 100 million ratings.

So half a million people, 20,000 movies, and, and ratings.

Our goal is to predict the, a few ratings for every user.

Where the evaluation criteria or the good, the way, how we

evaluate how good that our predictions, are based on the root mean squared error.

And, what we also know, for example, is that, that the root mean square error of

the Netflix system that was used in house and at that point in time was zero, 0.95.

So in some sense, on average this, this in,

in-house recommender system was about one star away from the, from the true rating.

So, this competition that was taking place in 2009 and 2010.

More than 2,700 teams joined the competition.

And what matrix also said was whoever is able to

improve their in-house recommended system by 10% gets $1 million prize.

字幕リスト動画再生

5 14 潜在因子モデル (5 14 Latent Factor Models 16 11)

recommend

critical

basically

average