Scribd
Upload
41 views8 pages

Matrix Factorization

The document discusses matrix factorization for recommender systems. Matrix factorization breaks down user preferences and item features into latent factors that can be used to predict unknown ratings. It explains how Netflix movie recommendations work by finding the dot product of a user's preference levels for different movie features and each movie's measured levels for those same features. Matrix factorization has the advantage of being able to predict ratings for movies a user has not yet seen based on their inferred preferences and the movies' inferred attributes.

Uploaded by

samuel.song
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
41 views8 pages

Matrix Factorization

The document discusses matrix factorization for recommender systems. Matrix factorization breaks down user preferences and item features into latent factors that can be used to predict unknown ratings. It explains how Netflix movie recommendations work by finding the dot product of a user's preference levels for different movie features and each movie's measured levels for those same features. Matrix factorization has the advantage of being able to predict ratings for movies a user has not yet seen based on their inferred preferences and the movies' inferred attributes.

Uploaded by

samuel.song
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

3/1/2021

Matrix Factorization

Matrix A
x
Matrix B
=

Matrix Factorization

User Preference
x
Item (movie) Features
=

1
3/1/2021

Netflix User-Movie Example

Movie Features
• Item Features = Product Attributes

Scary

2
3/1/2021

Movie Feature Level & User Preference Level

Movie Features

1 0 User preference level


User

Movie Movie feature level

Rating calculation using Dot Product


• The sum of the products of the corresponding entries of the
two sequences of numbers.
• Dot product of (1,0) and (3,1) = 1 x 3 + 0 x 1 = 3 + 0 = 3

3
3/1/2021

Dot Product
(1 x 3 + 0 x 1)
Rating =
=
User preference level
on each feature 1 0
X
Movie feature level

Dot Product

1 1

4
3/1/2021

Dot Product
Movie feature level User preference level

Rating calculation
• Identify (1) User preference level and (2) Movie feature level
User

User preference level

Feature

Movie feature level

Movie

5
3/1/2021

Rating calculation
• Calculate ratings using dot products.
User

User preference level

Feature Ratings
Movie feature level

Movie

Rating calculation Movie feature level

• Once (1) user preference level


and (2) movie feature level is
identified, (3) all the ratings
can be calculated using dot
User preference level

product.

6
3/1/2021

Advantage of Matrix Factorization


• In practice, since each user watched only a few movies among
all the Netflix movies, many ratings are empty (i.e. Sparse
Matrix)
• But, if there is good enough common behaviour, one can still
identify user preference level and item feature level.
• Then, one can do good enough guess for the ratings which is
unobserved yet.

Advantage of Matrix Factorization


All the rating can be
predicted using both
(1) Movie Feature
Levels and
(2) User Preference
Levels.

7
3/1/2021

Advantage of Matrix Factorization


- Predict the ratings in the white blank

Advantage of Matrix Factorization


- Predict the ratings in the white blank
- Which movie do you recommend for
User D?

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy