SlideShare a Scribd company logo
1 of 22
Download to read offline
Session-Based Recommender Systems
FREDERICK AYALA GÓMEZ
PHD STUDENT IN COMPUTER SCIENCE
@FREDAYALA FREDERICKAYALA
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 1
Presented at
Agenda
Why bother?
What are Recommender Systems?
How to build recommendations?
Hands-on GRU4Rec
Conclusions
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 2
Are we good at
choosing among
multiple
options?
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 3
happiness
choices
Having too many choices is stressful
- Takes time
- Increase uncertainty
Web and SoMe
as a channel to
consume
content,
products and
services.
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 4
Taken from https://www.wired.com/2004/10/tail/
What are
recommender
systems?
Not so formal definition
Finding
relevant things
for a user based
on some kind
of feedback
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 5
Products
Content
Services
Offer
User
interests
What is
Relevant
“Recommendations account for about 60% of all video
clicks from the home page” <- 2010James Davidson, Benjamin Liebald, Junning Liu, Palash Nandy, Taylor Van Vleet, Ullas Gargi, Sujoy Gupta, Yu He, Mike Lambert, Blake Livingston, and Dasarathi Sampath.
2010. The YouTube video recommendation system. In Proceedings of the fourth ACM conference on Recommender systems (RecSys '10). ACM, New York,
NY, USA, 293-296. DOI=http://dx.doi.org/10.1145/1864708.1864770
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 6
“Recommendations account for about 60% of all video clicks
from the home page.”
What are recommender systems?
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 7
Types of
Feedback
Explicit Feedback
“Asking the user”
Ratings
Like/Dislike
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 8
Implicit Feedback is
more common in
practice
Implicit
Feedback
“Observe the
user behaviors”
Clicks
Previously
seen items
How to build recommender systems?
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 9
Two books about Recommender Systems
Matrix Factorization
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 10
Represent the users/items feedback as a matrix. Then, find two lower dimension matrixes that factorize it
R
Usually very
sparse since
users consume
few items
ItemsUsers
≈ P
Q
Users
Itemsf
f
Y. Koren, R. Bell and C. Volinsky, "Matrix
Factorization Techniques for Recommender
Systems," in Computer, vol. 42, no. 8, pp. 30-37,
Aug. 2009.
doi: 10.1109/MC.2009.263
Minimize the squared error function
This is done usually by Stochastic Gradient Descent or Alternating Least Squares
For binary feedback use iALS
The predicted score for each user-item is given by
Works very well for personalization (e.g. newsletters)
No user profile
…
Old feedback
could be irrelevant
…
~84% of our
recommendations
are item-to-item
- DomonkosTikk,CEO@GravityR&D
Lessons learnt at building
recommendation services at industry
scale,
https://www.slideshare.net/domonkostikk/lessons-
learnt-at-building-recommendation-services-at-
industry-scale
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 11
Online Shopping Blogs
Browsing AnonymouslyNews/Expiring Content
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 12
Linden et al. 2003. “Amazon.com Recommendations: Item-to-Item Collaborative Filtering” IEEE Internet Computing
Davidson et al. 2010. “The YouTube video recommendation system”. ACM RecSys '10. BCN, Spain.
The next item is recommended based on the similarity between the actual item and other items.
The similarity is computed by using the two item-session vectors:
With this similarities we can find the k Nearest Neighbors of item i.
Usually a very strong baseline.
1 2 3 . . . . i . . . j . . . . . m
1 1 1
2
. 1 1
. 1 1
. 1 1
n 1
Items
Sessions
Item-KNN
“people who viewed this item also viewed…”
Item-to-Item Recommendations
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 13
User Item
User 1 A
User 1 B
User 1 C
User 2 B
User 2 C
User 2 D
User 3 B
User 3 C
User 3 A
Item i Item j Count
B C 3
A B 2
C A 1
C D 1
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 14
Where:
i: Current item
j: Next item
k: Other items in the dataset
yi: Latent Factor for item i
yj: Latent Factor for item j
yk: Latent Factor for item k
bj: Bias item j
bj: Bias item k
N. Koenigstein, Y. Koren. "Towards scalable and accurate item-oriented recommendations". ACM RecSys '13. NY, USA.
Learn model parameters that maximize the log-likelihood of
the training set:
Using Stochastic Gradient Descent and Importance
Sampling to reduce the complexity of the training
After concatenating the biases to the latent factors, the best next item j is the closest vector to the item i
Not the best baseline for top-K evaluations
Euclidean Item Recommender
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 15
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 16
Recurrent
Neural Networks
Train a model with a
sequence that predicts
a higher probability for
the next element.
Also called
sequence-to-sequence
models
We learned the alphabet in this direction
Telling it backwards is harder
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 17
Recurrent Neural Networks
Taken from: http://karpathy.github.io/2015/05/21/rnn-effectiveness/
H
E
L
O
H
E
L
O
Input and
outputs are
one-hot-
encoded
Backpropagate to fix
errors (i.e. move the
hidden layers weights a
small amount in the
direction of the loss
function gradient)
nok
nok ok
ok
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 18
GRU4RecB. Hidasi, et al. 2016 “Session-based recommendations with recurrent neural networks”. ICLR 16’, San Juan, PR
Taken from the authors ICLR poster http://www.hidasi.eu/content/gru4rec_iclr16_poster.pdf
Optimize directly to the ranking using BPR or TOP1
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 19
GRU4Rec - PerformanceB. Hidasi, et al. 2016 “Session-based recommendations with recurrent neural networks”. ICLR 16’, San Juan, PR
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 20
𝑅𝑒𝑐𝑎𝑙𝑙@20 =
# 𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛𝑠 𝑤𝑖𝑡ℎ 𝑡ℎ𝑒 𝑛𝑒𝑥𝑡 𝑑𝑒𝑠𝑖𝑟𝑒𝑑 𝑖𝑡𝑒𝑚 𝑖𝑛 𝑡ℎ𝑒 𝑇𝑜𝑝 20 𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑖𝑜𝑛𝑠
𝑇𝑜𝑡𝑎𝑙 # 𝑜𝑓 𝑟𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛𝑠
Collection of sessions – item click events
Hands-on GRU4REC
Jupyter Notebook at
https://github.com/frederickayala/session-based-recsys
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 21
Conclusions
4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 22
Frederick Ayala Gómez
PhD Student in Computer Science
@fredayala frederickayala
Session-Based Recommender Systems are very common in real life
GRU4Rec code is open and GPU accelerated with Theano
Thanks!
+12.8%

More Related Content

What's hot

Tutorial on People Recommendations in Social Networks - ACM RecSys 2013,Hong...
Tutorial on People Recommendations in Social Networks -  ACM RecSys 2013,Hong...Tutorial on People Recommendations in Social Networks -  ACM RecSys 2013,Hong...
Tutorial on People Recommendations in Social Networks - ACM RecSys 2013,Hong...
Anmol Bhasin
 
Recommendation engines
Recommendation enginesRecommendation engines
Recommendation engines
Georgian Micsa
 
Recommender system introduction
Recommender system   introductionRecommender system   introduction
Recommender system introduction
Liang Xiang
 

What's hot (20)

Recommendation System Explained
Recommendation System ExplainedRecommendation System Explained
Recommendation System Explained
 
Tutorial on People Recommendations in Social Networks - ACM RecSys 2013,Hong...
Tutorial on People Recommendations in Social Networks -  ACM RecSys 2013,Hong...Tutorial on People Recommendations in Social Networks -  ACM RecSys 2013,Hong...
Tutorial on People Recommendations in Social Networks - ACM RecSys 2013,Hong...
 
Recommender Systems
Recommender SystemsRecommender Systems
Recommender Systems
 
Recommender systems using collaborative filtering
Recommender systems using collaborative filteringRecommender systems using collaborative filtering
Recommender systems using collaborative filtering
 
Phd thesis final presentation
Phd thesis   final presentationPhd thesis   final presentation
Phd thesis final presentation
 
Recommender Systems (Machine Learning Summer School 2014 @ CMU)
Recommender Systems (Machine Learning Summer School 2014 @ CMU)Recommender Systems (Machine Learning Summer School 2014 @ CMU)
Recommender Systems (Machine Learning Summer School 2014 @ CMU)
 
Recommending What Video to Watch Next: A Multitask Ranking System
Recommending What Video to Watch Next: A Multitask Ranking SystemRecommending What Video to Watch Next: A Multitask Ranking System
Recommending What Video to Watch Next: A Multitask Ranking System
 
Recommendation system
Recommendation systemRecommendation system
Recommendation system
 
Vector database
Vector databaseVector database
Vector database
 
An introduction to Recommender Systems
An introduction to Recommender SystemsAn introduction to Recommender Systems
An introduction to Recommender Systems
 
Deep Learning for Recommender Systems
Deep Learning for Recommender SystemsDeep Learning for Recommender Systems
Deep Learning for Recommender Systems
 
Deep Natural Language Processing for Search and Recommender Systems
Deep Natural Language Processing for Search and Recommender SystemsDeep Natural Language Processing for Search and Recommender Systems
Deep Natural Language Processing for Search and Recommender Systems
 
Past, Present & Future of Recommender Systems: An Industry Perspective
Past, Present & Future of Recommender Systems: An Industry PerspectivePast, Present & Future of Recommender Systems: An Industry Perspective
Past, Present & Future of Recommender Systems: An Industry Perspective
 
Recommendation engines
Recommendation enginesRecommendation engines
Recommendation engines
 
Graph Neural Networks for Recommendations
Graph Neural Networks for RecommendationsGraph Neural Networks for Recommendations
Graph Neural Networks for Recommendations
 
Recommender system introduction
Recommender system   introductionRecommender system   introduction
Recommender system introduction
 
GRU4Rec v2 - Recurrent Neural Networks with Top-k Gains for Session-based Rec...
GRU4Rec v2 - Recurrent Neural Networks with Top-k Gains for Session-based Rec...GRU4Rec v2 - Recurrent Neural Networks with Top-k Gains for Session-based Rec...
GRU4Rec v2 - Recurrent Neural Networks with Top-k Gains for Session-based Rec...
 
Overview of recommender system
Overview of recommender systemOverview of recommender system
Overview of recommender system
 
Introduction to Graph Neural Networks: Basics and Applications - Katsuhiko Is...
Introduction to Graph Neural Networks: Basics and Applications - Katsuhiko Is...Introduction to Graph Neural Networks: Basics and Applications - Katsuhiko Is...
Introduction to Graph Neural Networks: Basics and Applications - Katsuhiko Is...
 
Recommendation Systems Basics
Recommendation Systems BasicsRecommendation Systems Basics
Recommendation Systems Basics
 

Similar to Session-Based Recommender Systems

Putting Data to Work: Moving science forward together beyond where we thought...
Putting Data to Work: Moving science forward together beyond where we thought...Putting Data to Work: Moving science forward together beyond where we thought...
Putting Data to Work: Moving science forward together beyond where we thought...
Erin Robinson
 
Data Presentation
Data PresentationData Presentation
Data Presentation
Jon Zurfluh
 
Confirming PagesLess managing. More teaching. Greater
Confirming PagesLess managing. More teaching. Greater Confirming PagesLess managing. More teaching. Greater
Confirming PagesLess managing. More teaching. Greater
AlleneMcclendon878
 
Recommender System in light of Big Data
Recommender System in light of Big DataRecommender System in light of Big Data
Recommender System in light of Big Data
Khadija Atiya
 

Similar to Session-Based Recommender Systems (20)

DataMind Pitch August 2013
DataMind Pitch August 2013DataMind Pitch August 2013
DataMind Pitch August 2013
 
Manage online information
Manage online informationManage online information
Manage online information
 
5 Benefits of Predictive Analytics for E-Commerce
5 Benefits of Predictive Analytics for E-Commerce5 Benefits of Predictive Analytics for E-Commerce
5 Benefits of Predictive Analytics for E-Commerce
 
Prospect for learning analytics to achieve adaptive learning model
Prospect for learning analytics to achieve adaptive learning modelProspect for learning analytics to achieve adaptive learning model
Prospect for learning analytics to achieve adaptive learning model
 
Learning Analytics for Adaptive Learning And Standardization
Learning Analytics for Adaptive Learning And StandardizationLearning Analytics for Adaptive Learning And Standardization
Learning Analytics for Adaptive Learning And Standardization
 
The Guzman Monthly, December 2015, v2 i12
The Guzman Monthly, December 2015, v2 i12The Guzman Monthly, December 2015, v2 i12
The Guzman Monthly, December 2015, v2 i12
 
Challenges in Analytics for BIG Data
Challenges in Analytics for BIG DataChallenges in Analytics for BIG Data
Challenges in Analytics for BIG Data
 
Putting Data to Work: Moving science forward together beyond where we thought...
Putting Data to Work: Moving science forward together beyond where we thought...Putting Data to Work: Moving science forward together beyond where we thought...
Putting Data to Work: Moving science forward together beyond where we thought...
 
Performance evaluation of Wearable Computing Frameworks
Performance evaluation of Wearable Computing FrameworksPerformance evaluation of Wearable Computing Frameworks
Performance evaluation of Wearable Computing Frameworks
 
Data Presentation
Data PresentationData Presentation
Data Presentation
 
AGU Leptoukh Lecture: Putting Data to Work: Moving science forward together b...
AGU Leptoukh Lecture: Putting Data to Work: Moving science forward together b...AGU Leptoukh Lecture: Putting Data to Work: Moving science forward together b...
AGU Leptoukh Lecture: Putting Data to Work: Moving science forward together b...
 
Demystifying Data Science with an introduction to Machine Learning
Demystifying Data Science with an introduction to Machine LearningDemystifying Data Science with an introduction to Machine Learning
Demystifying Data Science with an introduction to Machine Learning
 
E Marketing Week02
E Marketing Week02E Marketing Week02
E Marketing Week02
 
DRESD Project Presentation - December 2006
DRESD Project Presentation - December 2006DRESD Project Presentation - December 2006
DRESD Project Presentation - December 2006
 
Confirming PagesLess managing. More teaching. Greater
Confirming PagesLess managing. More teaching. Greater Confirming PagesLess managing. More teaching. Greater
Confirming PagesLess managing. More teaching. Greater
 
University Recommendation Support System using ML Algorithms
University Recommendation Support System using ML AlgorithmsUniversity Recommendation Support System using ML Algorithms
University Recommendation Support System using ML Algorithms
 
Linear Regression With R
Linear Regression With RLinear Regression With R
Linear Regression With R
 
Cse443 Project Report - LPU (Modern Big Data Analysis with SQL Specialization)
Cse443 Project Report - LPU (Modern Big Data Analysis with SQL Specialization)Cse443 Project Report - LPU (Modern Big Data Analysis with SQL Specialization)
Cse443 Project Report - LPU (Modern Big Data Analysis with SQL Specialization)
 
How Does Data Science Impact the Semantic Web?
How Does Data Science Impact the Semantic Web?How Does Data Science Impact the Semantic Web?
How Does Data Science Impact the Semantic Web?
 
Recommender System in light of Big Data
Recommender System in light of Big DataRecommender System in light of Big Data
Recommender System in light of Big Data
 

Recently uploaded

Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...
Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...
Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...
nirzagarg
 
Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...
Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...
Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...
HyderabadDolls
 
Top profile Call Girls In Latur [ 7014168258 ] Call Me For Genuine Models We ...
Top profile Call Girls In Latur [ 7014168258 ] Call Me For Genuine Models We ...Top profile Call Girls In Latur [ 7014168258 ] Call Me For Genuine Models We ...
Top profile Call Girls In Latur [ 7014168258 ] Call Me For Genuine Models We ...
gajnagarg
 
Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...
Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...
Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...
gajnagarg
 
Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...
Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...
Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...
nirzagarg
 
Top profile Call Girls In Chandrapur [ 7014168258 ] Call Me For Genuine Model...
Top profile Call Girls In Chandrapur [ 7014168258 ] Call Me For Genuine Model...Top profile Call Girls In Chandrapur [ 7014168258 ] Call Me For Genuine Model...
Top profile Call Girls In Chandrapur [ 7014168258 ] Call Me For Genuine Model...
gajnagarg
 
Computer science Sql cheat sheet.pdf.pdf
Computer science Sql cheat sheet.pdf.pdfComputer science Sql cheat sheet.pdf.pdf
Computer science Sql cheat sheet.pdf.pdf
SayantanBiswas37
 
Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...
Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...
Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...
gajnagarg
 
如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样
如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样
如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样
wsppdmt
 

Recently uploaded (20)

Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...
Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...
Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...
 
Top Call Girls in Balaghat 9332606886Call Girls Advance Cash On Delivery Ser...
Top Call Girls in Balaghat  9332606886Call Girls Advance Cash On Delivery Ser...Top Call Girls in Balaghat  9332606886Call Girls Advance Cash On Delivery Ser...
Top Call Girls in Balaghat 9332606886Call Girls Advance Cash On Delivery Ser...
 
Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...
Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...
Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...
 
Top profile Call Girls In Latur [ 7014168258 ] Call Me For Genuine Models We ...
Top profile Call Girls In Latur [ 7014168258 ] Call Me For Genuine Models We ...Top profile Call Girls In Latur [ 7014168258 ] Call Me For Genuine Models We ...
Top profile Call Girls In Latur [ 7014168258 ] Call Me For Genuine Models We ...
 
Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...
Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...
Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...
 
Gomti Nagar & best call girls in Lucknow | 9548273370 Independent Escorts & D...
Gomti Nagar & best call girls in Lucknow | 9548273370 Independent Escorts & D...Gomti Nagar & best call girls in Lucknow | 9548273370 Independent Escorts & D...
Gomti Nagar & best call girls in Lucknow | 9548273370 Independent Escorts & D...
 
Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...
Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...
Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...
 
TrafficWave Generator Will Instantly drive targeted and engaging traffic back...
TrafficWave Generator Will Instantly drive targeted and engaging traffic back...TrafficWave Generator Will Instantly drive targeted and engaging traffic back...
TrafficWave Generator Will Instantly drive targeted and engaging traffic back...
 
Predicting HDB Resale Prices - Conducting Linear Regression Analysis With Orange
Predicting HDB Resale Prices - Conducting Linear Regression Analysis With OrangePredicting HDB Resale Prices - Conducting Linear Regression Analysis With Orange
Predicting HDB Resale Prices - Conducting Linear Regression Analysis With Orange
 
High Profile Call Girls Service in Jalore { 9332606886 } VVIP NISHA Call Girl...
High Profile Call Girls Service in Jalore { 9332606886 } VVIP NISHA Call Girl...High Profile Call Girls Service in Jalore { 9332606886 } VVIP NISHA Call Girl...
High Profile Call Girls Service in Jalore { 9332606886 } VVIP NISHA Call Girl...
 
5CL-ADBA,5cladba, Chinese supplier, safety is guaranteed
5CL-ADBA,5cladba, Chinese supplier, safety is guaranteed5CL-ADBA,5cladba, Chinese supplier, safety is guaranteed
5CL-ADBA,5cladba, Chinese supplier, safety is guaranteed
 
Top profile Call Girls In Chandrapur [ 7014168258 ] Call Me For Genuine Model...
Top profile Call Girls In Chandrapur [ 7014168258 ] Call Me For Genuine Model...Top profile Call Girls In Chandrapur [ 7014168258 ] Call Me For Genuine Model...
Top profile Call Girls In Chandrapur [ 7014168258 ] Call Me For Genuine Model...
 
Dubai Call Girls Peeing O525547819 Call Girls Dubai
Dubai Call Girls Peeing O525547819 Call Girls DubaiDubai Call Girls Peeing O525547819 Call Girls Dubai
Dubai Call Girls Peeing O525547819 Call Girls Dubai
 
Gulbai Tekra * Cheap Call Girls In Ahmedabad Phone No 8005736733 Elite Escort...
Gulbai Tekra * Cheap Call Girls In Ahmedabad Phone No 8005736733 Elite Escort...Gulbai Tekra * Cheap Call Girls In Ahmedabad Phone No 8005736733 Elite Escort...
Gulbai Tekra * Cheap Call Girls In Ahmedabad Phone No 8005736733 Elite Escort...
 
Computer science Sql cheat sheet.pdf.pdf
Computer science Sql cheat sheet.pdf.pdfComputer science Sql cheat sheet.pdf.pdf
Computer science Sql cheat sheet.pdf.pdf
 
Vadodara 💋 Call Girl 7737669865 Call Girls in Vadodara Escort service book now
Vadodara 💋 Call Girl 7737669865 Call Girls in Vadodara Escort service book nowVadodara 💋 Call Girl 7737669865 Call Girls in Vadodara Escort service book now
Vadodara 💋 Call Girl 7737669865 Call Girls in Vadodara Escort service book now
 
Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...
Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...
Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...
 
Digital Advertising Lecture for Advanced Digital & Social Media Strategy at U...
Digital Advertising Lecture for Advanced Digital & Social Media Strategy at U...Digital Advertising Lecture for Advanced Digital & Social Media Strategy at U...
Digital Advertising Lecture for Advanced Digital & Social Media Strategy at U...
 
Discover Why Less is More in B2B Research
Discover Why Less is More in B2B ResearchDiscover Why Less is More in B2B Research
Discover Why Less is More in B2B Research
 
如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样
如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样
如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样
 

Session-Based Recommender Systems

  • 1. Session-Based Recommender Systems FREDERICK AYALA GÓMEZ PHD STUDENT IN COMPUTER SCIENCE @FREDAYALA FREDERICKAYALA 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 1 Presented at
  • 2. Agenda Why bother? What are Recommender Systems? How to build recommendations? Hands-on GRU4Rec Conclusions 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 2
  • 3. Are we good at choosing among multiple options? 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 3 happiness choices Having too many choices is stressful - Takes time - Increase uncertainty
  • 4. Web and SoMe as a channel to consume content, products and services. 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 4 Taken from https://www.wired.com/2004/10/tail/
  • 5. What are recommender systems? Not so formal definition Finding relevant things for a user based on some kind of feedback 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 5 Products Content Services Offer User interests What is Relevant
  • 6. “Recommendations account for about 60% of all video clicks from the home page” <- 2010James Davidson, Benjamin Liebald, Junning Liu, Palash Nandy, Taylor Van Vleet, Ullas Gargi, Sujoy Gupta, Yu He, Mike Lambert, Blake Livingston, and Dasarathi Sampath. 2010. The YouTube video recommendation system. In Proceedings of the fourth ACM conference on Recommender systems (RecSys '10). ACM, New York, NY, USA, 293-296. DOI=http://dx.doi.org/10.1145/1864708.1864770 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 6 “Recommendations account for about 60% of all video clicks from the home page.”
  • 7. What are recommender systems? 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 7
  • 8. Types of Feedback Explicit Feedback “Asking the user” Ratings Like/Dislike 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 8 Implicit Feedback is more common in practice Implicit Feedback “Observe the user behaviors” Clicks Previously seen items
  • 9. How to build recommender systems? 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 9 Two books about Recommender Systems
  • 10. Matrix Factorization 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 10 Represent the users/items feedback as a matrix. Then, find two lower dimension matrixes that factorize it R Usually very sparse since users consume few items ItemsUsers ≈ P Q Users Itemsf f Y. Koren, R. Bell and C. Volinsky, "Matrix Factorization Techniques for Recommender Systems," in Computer, vol. 42, no. 8, pp. 30-37, Aug. 2009. doi: 10.1109/MC.2009.263 Minimize the squared error function This is done usually by Stochastic Gradient Descent or Alternating Least Squares For binary feedback use iALS The predicted score for each user-item is given by Works very well for personalization (e.g. newsletters)
  • 11. No user profile … Old feedback could be irrelevant … ~84% of our recommendations are item-to-item - DomonkosTikk,CEO@GravityR&D Lessons learnt at building recommendation services at industry scale, https://www.slideshare.net/domonkostikk/lessons- learnt-at-building-recommendation-services-at- industry-scale 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 11 Online Shopping Blogs Browsing AnonymouslyNews/Expiring Content
  • 12. 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 12 Linden et al. 2003. “Amazon.com Recommendations: Item-to-Item Collaborative Filtering” IEEE Internet Computing Davidson et al. 2010. “The YouTube video recommendation system”. ACM RecSys '10. BCN, Spain. The next item is recommended based on the similarity between the actual item and other items. The similarity is computed by using the two item-session vectors: With this similarities we can find the k Nearest Neighbors of item i. Usually a very strong baseline. 1 2 3 . . . . i . . . j . . . . . m 1 1 1 2 . 1 1 . 1 1 . 1 1 n 1 Items Sessions Item-KNN “people who viewed this item also viewed…”
  • 13. Item-to-Item Recommendations 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 13 User Item User 1 A User 1 B User 1 C User 2 B User 2 C User 2 D User 3 B User 3 C User 3 A Item i Item j Count B C 3 A B 2 C A 1 C D 1
  • 14. 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 14 Where: i: Current item j: Next item k: Other items in the dataset yi: Latent Factor for item i yj: Latent Factor for item j yk: Latent Factor for item k bj: Bias item j bj: Bias item k N. Koenigstein, Y. Koren. "Towards scalable and accurate item-oriented recommendations". ACM RecSys '13. NY, USA. Learn model parameters that maximize the log-likelihood of the training set: Using Stochastic Gradient Descent and Importance Sampling to reduce the complexity of the training After concatenating the biases to the latent factors, the best next item j is the closest vector to the item i Not the best baseline for top-K evaluations Euclidean Item Recommender
  • 16. 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 16 Recurrent Neural Networks Train a model with a sequence that predicts a higher probability for the next element. Also called sequence-to-sequence models We learned the alphabet in this direction Telling it backwards is harder
  • 17. 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 17 Recurrent Neural Networks Taken from: http://karpathy.github.io/2015/05/21/rnn-effectiveness/ H E L O H E L O Input and outputs are one-hot- encoded Backpropagate to fix errors (i.e. move the hidden layers weights a small amount in the direction of the loss function gradient) nok nok ok ok
  • 19. GRU4RecB. Hidasi, et al. 2016 “Session-based recommendations with recurrent neural networks”. ICLR 16’, San Juan, PR Taken from the authors ICLR poster http://www.hidasi.eu/content/gru4rec_iclr16_poster.pdf Optimize directly to the ranking using BPR or TOP1 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 19
  • 20. GRU4Rec - PerformanceB. Hidasi, et al. 2016 “Session-based recommendations with recurrent neural networks”. ICLR 16’, San Juan, PR 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 20 𝑅𝑒𝑐𝑎𝑙𝑙@20 = # 𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛𝑠 𝑤𝑖𝑡ℎ 𝑡ℎ𝑒 𝑛𝑒𝑥𝑡 𝑑𝑒𝑠𝑖𝑟𝑒𝑑 𝑖𝑡𝑒𝑚 𝑖𝑛 𝑡ℎ𝑒 𝑇𝑜𝑝 20 𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑖𝑜𝑛𝑠 𝑇𝑜𝑡𝑎𝑙 # 𝑜𝑓 𝑟𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛𝑠 Collection of sessions – item click events
  • 21. Hands-on GRU4REC Jupyter Notebook at https://github.com/frederickayala/session-based-recsys 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 21
  • 22. Conclusions 4/20/2017 @FREDAYALA LINKEDIN.COM/IN/FREDERICKAYALA 22 Frederick Ayala Gómez PhD Student in Computer Science @fredayala frederickayala Session-Based Recommender Systems are very common in real life GRU4Rec code is open and GPU accelerated with Theano Thanks! +12.8%