Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender...
Transcript of Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender...
![Page 1: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/1.jpg)
Online Recommender Systems
CS 780/880
![Page 2: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/2.jpg)
Outline
I Content-based o�er recommendations
I Collaborative filteringI Recommender systemsI Online travel recommendationsI Collaborative filteringI Matrix factorization and completion
![Page 3: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/3.jpg)
Outline
I Content-based o�er recommendations
I Collaborative filteringI Recommender systemsI Online travel recommendationsI Collaborative filteringI Matrix factorization and completion
![Page 4: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/4.jpg)
Example recommender systems
I Movie recommendations (Netflix)I Personalized music (Pandora)I Relevant product recommendations (Amazon)I Online advertising (Facebook)I Online dating (OK Cupid)I Special o�ers, coupons, and discounts (Stores)
![Page 5: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/5.jpg)
Google Play recommendations
![Page 6: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/6.jpg)
Amazon recommendations
![Page 7: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/7.jpg)
IBM travel recommendations
I Recommend relevant products on website of a tour operator
![Page 8: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/8.jpg)
Travel recommendation steps
Unknown customer
Browsing history (trips)
Best guess Recommended trips
![Page 9: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/9.jpg)
Travel recommendation steps
Unknown customer
Browsing history (trips)
Best guess Recommended trips
![Page 10: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/10.jpg)
Travel recommendation steps
Unknown customer
Browsing history (trips)
Best guess
Recommended trips
![Page 11: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/11.jpg)
Travel recommendation steps
Unknown customer
Browsing history (trips)
Best guess Recommended trips
![Page 12: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/12.jpg)
Benefits of online recommendations
I What are the benefits?I For users/customer:
1. Find the right product or information2. See only relevant advertising3. Discover diverse products or information sources
I For businesses/websites:1. Increase user satisfaction2. Sell more products and right products3. Discriminate between customers4. Predict and understand customer preferences
![Page 13: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/13.jpg)
Benefits of online recommendations
I What are the benefits?I For users/customer:
1. Find the right product or information2. See only relevant advertising3. Discover diverse products or information sources
I For businesses/websites:1. Increase user satisfaction2. Sell more products and right products3. Discriminate between customers4. Predict and understand customer preferences
![Page 14: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/14.jpg)
Benefits of online recommendations
I What are the benefits?I For users/customer:
1. Find the right product or information2. See only relevant advertising3. Discover diverse products or information sources
I For businesses/websites:1. Increase user satisfaction2. Sell more products and right products3. Discriminate between customers4. Predict and understand customer preferences
![Page 15: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/15.jpg)
Types of recommender systems
I By delivery:I Item-to-item recommendations: e.g. Amazon productsI User-to-item recommendations: e.g. Netflix, Google Play
I By information used:I Content-based: Use item description and user profile
I Use when rich user profile and content information is availableI Collaborative filtering: Use preferences of other users
I When rich interaction history is availableI Hybrid: Combination
![Page 16: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/16.jpg)
Types of recommender systems
I By delivery:I Item-to-item recommendations: e.g. Amazon productsI User-to-item recommendations: e.g. Netflix, Google Play
I By information used:I Content-based: Use item description and user profile
I Use when rich user profile and content information is availableI Collaborative filtering: Use preferences of other users
I When rich interaction history is availableI Hybrid: Combination
![Page 17: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/17.jpg)
Types of recommender systems
I By delivery:I Item-to-item recommendations: e.g. Amazon productsI User-to-item recommendations: e.g. Netflix, Google Play
I By information used:I Content-based: Use item description and user profile
I Use when rich user profile and content information is availableI Collaborative filtering: Use preferences of other users
I When rich interaction history is availableI Hybrid: Combination
![Page 18: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/18.jpg)
Types of recommender systems
I By delivery:I Item-to-item recommendations: e.g. Amazon productsI User-to-item recommendations: e.g. Netflix, Google Play
I By information used:I Content-based: Use item description and user profile
I Use when rich user profile and content information is availableI Collaborative filtering: Use preferences of other users
I When rich interaction history is availableI Hybrid: Combination
![Page 19: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/19.jpg)
Types of recommender systems
I By delivery:I Item-to-item recommendations: e.g. Amazon productsI User-to-item recommendations: e.g. Netflix, Google Play
I By information used:I Content-based: Use item description and user profile
I Use when rich user profile and content information is availableI Collaborative filtering: Use preferences of other users
I When rich interaction history is availableI Hybrid: Combination
![Page 20: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/20.jpg)
Content-based or collaborative filtering?
Unknown customer
Browsing history (trips)
Best guess Recommended trips
![Page 21: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/21.jpg)
Content-based recommender systems
I Learn user preference model based on a�ributesI Use historical user preference data
User MovieGender Age Genre Year Rating
Male 16 Comedy 1998 1Male 98 Horror 1912 5Female 43 Action 2016 3
I Use supervised learning methods
![Page 22: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/22.jpg)
Collaborative-filtering recommender systems
I Very flexible: no need to know content or user profilesI Simple and powerful methodsI Training data: Partial user-item preferences
3 ? ? 1 3 1 1 5 ? 5
? 2 3 ? 3 ? 2 ? ? ?
? 5 ? ? 1 ? ? 2 ? 4
4 3 ? 3 ? ? ? ? ? ?
2 3 5 2 ? ? ? 5 ? 4
? ? 1 2 5 3 4 ? ? ?
useritem
I Making recommendations: Fill in the blanks (?)
![Page 23: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/23.jpg)
Nearest neighbors: Use similar users
I Infer preferences from similar usersI Other users:
3 ? ? 1 3 1 1 5 ? 5
4 3 ? 3 ? ? ? ? 5 ?
user
item
I Current user:
? 3 3 3 ? ? 1 5 ? ?
( )
user
I Basic algorithm:1. Find similar user (e.g. 2 ratings same)2. Infer unknown preferences
I Why does this fail?
![Page 24: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/24.jpg)
Nearest neighbors: Use similar users
I Infer preferences from similar usersI Other users:
3 ? ? 1 3 1 1 5 ? 5
4 3 ? 3 ? ? ? ? 5 ?
user
item
I Current user:
? 3 3 3 ? ? 1 5 ? ?
( )
user
I Basic algorithm:1. Find similar user (e.g. 2 ratings same)2. Infer unknown preferences
I Why does this fail?
![Page 25: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/25.jpg)
Nearest neighbors: Use similar users
I Infer preferences from similar usersI Other users:
3 ? ? 1 3 1 1 5 ? 5
4 3 ? 3 ? ? ? ? 5 ?
user
item
I Current user:
? 3 3 3 ? ? 1 5 ? ?
( )
user
I Basic algorithm:1. Find similar user (e.g. 2 ratings same)2. Infer unknown preferences
I Why does this fail?
![Page 26: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/26.jpg)
Nearest neighbors: Use similar users
I Infer preferences from similar usersI Other users:
3 ? ? 1 3 1 1 5 ? 5
4 3 ? 3 ? ? ? ? 5 ?
user
item
I Current user:
4 3 3 3 ? ? 1 5 5 ?
( )
user
I Basic algorithm:1. Find similar user (e.g. 2 ratings same)2. Infer unknown preferences
I Why does this fail?
![Page 27: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/27.jpg)
Nearest neighbors: Use similar users
I Infer preferences from similar usersI Other users:
3 ? ? 1 3 1 1 5 ? 5
4 3 ? 3 ? ? ? ? 5 ?
user
item
I Current user:
4 3 3 3 ? ? 1 5 5 ?
( )
user
I Basic algorithm:1. Find similar user (e.g. 2 ratings same)2. Infer unknown preferences
I Why does this fail?
![Page 28: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/28.jpg)
Nearest neighbors: Use similar users
I Infer preferences from similar usersI Other users:
3 ? ? 1 3 1 1 5 ? 5
4 3 ? 3 ? ? ? ? 5 ?
user
item
I Current user:
4 3 3 3 ? ? 1 5 5 ?
( )
user
I Basic algorithm:1. Find similar user (e.g. 2 ratings same)2. Infer unknown preferences
I Why does this fail? Conflicting preferences!
![Page 29: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/29.jpg)
Matrix factorizationI Be�er model for addressing incompatible preferencesI Assumption: Latent (unobserved) customer type, e.g.
1. Adventure seeking2. Luxury oriented
I Low-rank matrix decomposition:
2 3 3 1 1
2 3 3 1 1
4 3 1 1 3
4 3 1 1 3
4 3 1 1 3
user
item
=
0 1
0 1
1 0
1 0
1 0
}
user
type
4 3 1 1 3
2 3 3 1 1
( )
type
item
D U QT=
I D11 = U1QT1
I Not integral in general!
![Page 30: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/30.jpg)
Matrix factorizationI Be�er model for addressing incompatible preferencesI Assumption: Latent (unobserved) customer type, e.g.
1. Adventure seeking2. Luxury oriented
I Low-rank matrix decomposition:
2 3 3 1 1
2 3 3 1 1
4 3 1 1 3
4 3 1 1 3
4 3 1 1 3
user
item
=
0 1
0 1
1 0
1 0
1 0
}
user
type
4 3 1 1 3
2 3 3 1 1
( )
type
item
D U QT=
I D11 = U1QT1
I Not integral in general!
![Page 31: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/31.jpg)
Computing matrix factorization
I Given a small constant kI How to compute the best factorization U,Q?
D?= UQT s.t. rank(U) = rank(Q) = k
I Solve optimization problem:
minU,Q‖D − UQT‖2F s.t. rank(U) = rank(Q) = k
I Frobenius norm: ‖A‖2F =∑
i,j A2ij =
∑i σ
2i
I Solve optimally by SVD when D is completeI NP-hard when D is incompleteI Simple, practical, and e�ective method
1. Fix U and solve: minQ ‖D − UQT‖2F s.t. rank(Q) = k2. Fix Q and solve: minU ‖D − UQT‖2F s.t. rank(U) = k
![Page 32: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/32.jpg)
Computing matrix factorization
I Given a small constant kI How to compute the best factorization U,Q?
D?= UQT s.t. rank(U) = rank(Q) = k
I Solve optimization problem:
minU,Q‖D − UQT‖2F s.t. rank(U) = rank(Q) = k
I Frobenius norm: ‖A‖2F =∑
i,j A2ij =
∑i σ
2i
I Solve optimally by SVD when D is completeI NP-hard when D is incompleteI Simple, practical, and e�ective method
1. Fix U and solve: minQ ‖D − UQT‖2F s.t. rank(Q) = k2. Fix Q and solve: minU ‖D − UQT‖2F s.t. rank(U) = k
![Page 33: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/33.jpg)
Matrix factorization: Results
I One year worth of website click-stream dataI Dependence on number of previously visited sitesI Matrix factorization (green) vs baseline (blue)
![Page 34: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/34.jpg)
Beyond simple models: Dynamic user behavior
I Hidden Markov Model
I Strategic recommendations: optimal in long run(conversion/satisfaction)
![Page 35: Online Recommender Systems - Computer Sciencempetrik/teaching/intro_ml...Types of recommender systems I By delivery: I Item-to-item recommendations: e.g. Amazon products I User-to-item](https://reader035.fdocuments.net/reader035/viewer/2022071405/60fa76484636b80d7430b428/html5/thumbnails/35.jpg)
Recommendations on social media (Twi�er)
I Challenges1. Large volume: (6000 t/s)2. Short (cryptic) text:
“Cruise with me on thiswild ride”
I Approach:1. Use language models to
identify travel intent2. Matrix factorization to
match catalogdescriptions with tweets
I Positive engagement: 15%