lessons from the netflix prize robert bell at&t labs-research in collaboration with chris...
TRANSCRIPT
Lessons fromthe Netflix Prize
Robert Bell AT&T Labs-Research
In collaboration with
Chris Volinsky, AT&T Labs-Research & Yehuda Koren, Yahoo! Research
2
“We’re quite curious, really. To the tune of one million dollars.” – Netflix Prize rules
• Goal to improve on Netflix’s existing movie recommendation technology
• Contest began October 2, 2006• Prize
– Based on reduction in root mean squared error (RMSE) on test data
– $1,000,000 grand prize for 10% drop (19% for MSE)– Or, $50,000 progress for best result each year
3
Data Details
• Training data– 100 million ratings (from 1 to 5 stars)– 6 years (2000-2005)– 480,000 users– 17,770 “movies”
• Test data– Last few ratings of each user– Split as shown on next slide
4
Test Data Split into Three Pieces
• Probe– Ratings released– Allows participants to assess
methods directly
• Daily submissions allowed for combined Quiz/Test data– Identity of Quiz cases
withheld– RMSE released for Quiz– Test RMSE withheld– Prizes based on Test RMSE
5
Higher Mean Rating in Probe Data
0
5
10
15
20
25
30
35
40
1 2 3 4 5
Rating
Per
cen
tag
e
Training (m = 3.60)
Probe (m = 3.67)
6
2004
Something Happened in Early 2004
Data about the MoviesMost Loved Movies Avg rating Count
The Shawshank Redemption 4.593 137812
Lord of the Rings: The Return of the King 4.545 133597
The Green Mile 4.306 180883
Lord of the Rings: The Two Towers 4.460 150676
Finding Nemo 4.415 139050
Raiders of the Lost Ark 4.504 117456
Most Rated MoviesMiss Congeniality
Independence Day
The Patriot
The Day After Tomorrow
Pretty Woman
Pirates of the Caribbean
Highest VarianceThe Royal Tenenbaums
Lost In Translation
Pearl Harbor
Miss Congeniality
Napolean Dynamite
Fahrenheit 9/11
8
Most Active Users
User ID # Ratings Mean Rating
305344 17,651 1.90
387418 17,432 1.81
2439493 16,560 1.22
1664010 15,811 4.26
2118461 14,829 4.08
1461435 9,820 1.37
1639792 9,764 1.33
1314869 9,739 2.95
9
Major Challenges
1. Size of data– Places premium on efficient algorithms– Stretched memory limits of standard PCs
2. 99% of data are missing– Eliminates many standard prediction methods– Certainly not missing at random
3. Training and test data differ systematically– Test ratings are later– Test cases are spread uniformly across users
10
Major Challenges (cont.)
4. Countless factors may affect ratings– Genre, movie/TV series/other– Style of action, dialogue, plot, music et al.– Director, actors– Rater’s mood
5. Large imbalance in training data– Number of ratings per user or movie varies by
several orders of magnitude– Information to estimate individual parameters varies
widely
11
Ratings per Movie in Training Data
Avg #ratings/movie: 5627
12
Ratings per User in Training Data
Avg #ratings/user: 208
13
The Fundamental Challenge
• How can we estimate as much signal as possible where there are sufficient data, without over fitting where data are scarce?
14
Recommender Systems
• Personalized recommendations of items (e.g., movies) to users
• Increasingly common– To deal with explosive number of choices on
the internet– Netflix– Amazon– Many others
15
Content Based Systems
• A pre-specified list of attributes
• Score each item on all attributes
• User interest obtained for the same attributes– Direct solicitation, or– Estimated based on user rating, purchases,
or other behavior
16
Pandora
• Music recommendation system
• Songs rated on 400+ attributes– Music genome project– Roots, instrumentation, lyrics, vocals
• Two types of user feedback– Seed songs– Thumbs up/down for recommended songs
17
Collaborative Filtering (CF)
• Avoids need for:– Determining “proper” content– Collecting information about items or users
• Infers user-item relationships from purchases or ratings
• Used by Amazon and Netflix• Two main CF tools
– Nearest neighbors– Latent factor models
18
Nearest Neighbor Methods
• Most common CF tool at the beginning of the contest• Predict rating for a specific user-item pair based on
ratings of– Similar items– By the same user– Or vice versa
•
• Pearson correlation or cosine similarity
);(
);(ˆuiNj ij
uiNj ujij
ui s
rsr
19
Merits of Nearest Neighbors
• Few modeling assumptions• Few tuning parameters to learn• Easy to explain to users
– Dear Amazon.com Customer, We've noticed that customers who have purchased or rated How Does the Show Go On: An Introduction to the Theater by Thomas Schumacher have also purchased Princess Protection Program #1: A Royal Makeover (Disney Early Readers).
20
Latent Factor Models
• Models with latent classes of items and users– Individual items and users are assigned to either a
single class or a mixture of classes
• Neural networks– Restricted Boltzmann machines
• Singular Value Decomposition (SVD)– AKA matrix factorization– Items and users described by unobserved factors– Main method used by leaders of competition
21
SVD
• Dimension reduction technique for matrices• Each item summarized by a
d-dimensional vector qi • Similarly, each user summarized by pu
• Choose d much smaller than number of items or users– e.g., d = 50 << 18,000 or 480,000
• Predicted rating for Item i by User u– Inner product of qi and pu
– ˆor ˆ ''uiiuuiuiui pqbarpqr
22
Geared towards females
Geared towards males
serious
escapist
The PrincessDiaries
The Lion King
Braveheart
Lethal Weapon
Independence Day
AmadeusThe Color Purple
Dumb and Dumber
Ocean’s 11Sense and Sensibility
23
Geared towards females
Geared towards males
serious
escapist
The PrincessDiaries
The Lion King
Braveheart
Lethal Weapon
Independence Day
AmadeusThe Color Purple
Dumb and Dumber
Ocean’s 11Sense and Sensibility
Gus
Dave
24
Regularization for SVD
• Want to minimize SSE for Test data• One idea: Minimize SSE for Training data
– Want large d to capture all the signals– But, Test RMSE begins to rise for d > 2
• Regularization is needed– Allow rich model where there are sufficient data– Shrink aggressively where data are scarce
• Minimize
ii
uu
trainingiuui qpqpr
222' )(
25
Geared towards females
Geared towards males
serious
escapist
The PrincessDiaries
The Lion King
Braveheart
Lethal Weapon
Independence Day
AmadeusThe Color Purple
Dumb and Dumber
Ocean’s 11Sense and Sensibility
Gus
26
Geared towards females
Geared towards males
serious
escapist
The PrincessDiaries
The Lion King
Braveheart
Lethal Weapon
Independence Day
AmadeusThe Color Purple
Dumb and Dumber
Ocean’s 11Sense and Sensibility
Gus
27
Geared towards females
Geared towards males
serious
escapist
The PrincessDiaries
The Lion King
Braveheart
Lethal Weapon
Independence Day
AmadeusThe Color Purple
Dumb and Dumber
Ocean’s 11Sense and Sensibility
Gus
28
Geared towards females
Geared towards males
serious
escapist
The PrincessDiaries
The Lion King
Braveheart
Lethal Weapon
Independence Day
AmadeusThe Color Purple
Dumb and Dumber
Ocean’s 11Sense and Sensibility
Gus
29
Estimation for SVD
• Fit by gradient descent– Loop over observed ratings– Update each relevant parameter– Small step in each parameter, proportional to gradient– Repeat until convergence
• Alternatively, fit by sequence of ridge regressions– Fix item factors– Loop over users, estimating user factors– Do same to estimate item factors– Repeat until convergence
Improvements toCollaborative Filtering
• Fine tune existing methods
• Incorporate alternative “effects”
• Incorporate a variety of modeling methods
• Careful regularization to avoid over fitting
Localized SVD
• SVD uses all of a user’s ratings to train the user’s factors
• But what if the user is multiple people?– Different factor values may apply to movies rated by
Mom vs. Dad vs. the Kids
• This approach computes user factors, pu , specific to the movie being predicted– Given all the {qi}, pu is the solution of a ridge regression
– Weighted ridge regressions with higher weights for movies similar to the target movie
Improvement from Localized SVD
33
Lesson #1: Data >> Models
• Very limited feature set– User, movie, date– Places focus on models/algorithms
• Major steps forward associated with incorporating new data features– What movies a user rated– Temporal effects
34
You are What You Rate
• What you rate (and don’t) provides information about your preferences
• Paterek’s NSVD explicitly characterizes users by which movies they like
• Incorporate what a user rated into the user factor–
• Substantially reduces RMSE
ˆ)(
21'
uNj
j/-
uiiuui y|N(u)|pqbar
35
Temporal Effects
• User behavior may change over time– Ratings go up or down– Interests change– For example, with addition of a new rater
• Allow user biases and/or factors to change over time– – Model au(t) and pu(t) as linear, unrestricted,
or a sum of both types
)()()()(ˆ)(
21'
uNj
j/-
uiiuui y|N(u)|tpqtbtatr
3636
Geared towards females
Geared towards males
serious
escapist
The PrincessDiaries
The Lion King
Braveheart
Lethal Weapon
Independence Day
AmadeusThe Color Purple
Dumb and Dumber
Ocean’s 11Sense and Sensibility
Gus
3737
Geared towards females
Geared towards males
serious
escapist
The PrincessDiaries
The Lion King
Braveheart
Lethal Weapon
Independence Day
AmadeusThe Color Purple
Dumb and Dumber
Ocean’s 11Sense and Sensibility
Gus
3838
Geared towards females
Geared towards males
serious
escapist
The PrincessDiaries
The Lion King
Braveheart
Lethal Weapon
Independence Day
AmadeusThe Color Purple
Dumb and Dumber
Ocean’s 11Sense and Sensibility
Gus
3939
Geared towards females
Geared towards males
serious
escapist
The PrincessDiaries
The Lion King
Braveheart
Lethal Weapon
Independence Day
AmadeusThe Color Purple
Dumb and Dumber
Ocean’s 11Sense and Sensibility
Gus +
40
#2: The Power of Regularized SVD Fit by Gradient Descent
• Allowed anyone to approach early leaders– Powerful predictor– Efficient– Easy to program
• Flexibility to incorporate additional features– Implicit feedback– Temporal effects– Neighborhood effects
• Accurate regularization is essential
41
Factor models: RMSE vs. #parameters
200
100
50
200100
50
500200
100
50
500200100
15001000500200100
50
0.875
0.880
0.885
0.890
0.895
0.900
0.905
10 100 1000 10000 100000
Millions of Parameters
RMSE
Basic SVD
… + What was Rated
… + Linear Time Factors
… + Per-Day User Biases
… + per-Day User Factors
#3: The Wisdom of Crowds (of Models)
• All models are wrong; some are useful – G. Box
• Used linear blends of many prediction sets– 107 in Year 1– Over 800 at the end
• Difficult, or impossible, to build the grand unified model
• Mega blends are not needed in practice– A handful of simple models achieves 80 percent of
the improvement of the full blend
43
#4: Find Good Teammates
• Yehuda Koren– The engine of progress for the Netflix Prize– Implicit feedback– Temporal effects– Nearest neighbor modeling
• Big Chaos: Michael Jahrer, Andreas Toscher (Year 2)– Optimization of tuning parameters– Blending methods
• Pragmatic Theory: Martin Chabbert, Martin Piotte (Year 3)– Some movies age better than others– Link functions
44
The Final Leaderboard
45
Test Set Results
• The Ensemble: 0.856714
46
Test Set Results
• The Ensemble: 0.856714
• BellKor’s Pragmatic Theory: 0.856704
47
Test Set Results
• The Ensemble: 0.856714
• BellKor’s Pragmatic Theory: 0.856704
• Both scores round to 0.8567
48
Test Set Results
• The Ensemble: 0.856714
• BellKor’s Pragmatic Theory: 0.856704
• Both scores round to 0.8567
• Tie breaker is submission date/time
49
Final Test Set Leaderboard
Who Got the Money?
• AT&T’s donated its full share to organizations supporting science education
• Young Science Achievers Program• New Jersey Institute of Technology pre-college
and educational opportunity programs• North Jersey Regional Science Fair• Neighborhoods Focused on African American
Youth
51
#5: Is This the Way to Do Science?
• Big Success for Netflix– Lots of cheap labor, good publicity– Already incorporated 6 percent improvement– Potential for much more using other data they have
• Big advances to the science of recommender systems– Regularized SVD– Identification of new features– Understanding nearest neighbors– Contributions to literature
52
Why Did this Work so Well?
• Industrial strength data
• Very good design
• Accessibility to anyone with a PC
• Free flow of ideas– Leaderboard– Forum– Workshop and papers
• Money?
53
But There are Limitations
• Need a conceptually simple task
• Winner-take-all has drawbacks
• Intellectual property and liability issues
• How many prizes can overlap?
54
Thank You!
• www.netflixprize.com– …/leaderboard– …/community
• Click BellKor’s Pragmatic Chaos or The Ensemble on Leaderboard for details