Rethinking Collabora0ve Filtering: A Prac0cal Perspec0ve on - - PowerPoint PPT Presentation

▶
rethinking collabora0ve filtering a prac0cal perspec0ve
SMART_READER_LITE
LIVE PREVIEW

Rethinking Collabora0ve Filtering: A Prac0cal Perspec0ve on - - PowerPoint PPT Presentation

Jul 30, 2023 193 likes •469 views

Rethinking Collabora0ve Filtering: A Prac0cal Perspec0ve on State-Of-The- Art Research Based on Real-World Insights and Challenges Noam Koenigstein 1 RECOMMENDATIONS IN MICROSOFT STORE 2 Windows Store 3 The Xbox Marketplace Xbox

slide-1
SLIDE 1

Rethinking Collabora0ve Filtering: A Prac0cal Perspec0ve on State-Of-The- Art Research Based on “Real-World” Insights and Challenges

Noam Koenigstein

1

slide-2
SLIDE 2

RECOMMENDATIONS IN MICROSOFT STORE

2

slide-3
SLIDE 3

3

Windows Store

slide-4
SLIDE 4

Xbox Marketplace

4

The Xbox Marketplace

slide-5
SLIDE 5

5

Groove Radio

slide-6
SLIDE 6

MicrosoE’s Web-Store

6

slide-7
SLIDE 7

A DECADE AGO…

7

slide-8
SLIDE 8

A Decade Ago… The NeJlix Prize

The goal: 10% improvement in RMSE over NeSlix’s Cinematch It took tens of thousands of par0cipants over 2 years….

8

𝑆𝑁𝑇𝐹=√⁠​1/𝑜 ∑𝑗=1↑𝑜▒​(​𝑧↓𝑗 −​ 𝑧 ↓𝑗 )↑2

slide-9
SLIDE 9

The Problem with RaMngs

  • They do not exist!
  • Missing items not at random

CollaboraMve Filtering and the Missing at Random AssumpMon

  • B. M. Marlin, R. S. Zemel, S. Roweis, M. Slaney
  • Ra0ngs are fuzzy and influenced by the order of items

RaMng vs. Preference: A comparaMve study of self-reporMng

  • G. N. Yannakakis, J. Hallam
  • Learning ra0ngs is very different from personaliza0on!

Yahoo! Music RecommendaMons: Modeling Music RaMngs with Temporal Dynamics and Taxonomy Gideon Dror, Noam Koenigstein and Yehuda Koren

9

slide-10
SLIDE 10

IF NOT RMSE THEN WHAT?

10

slide-11
SLIDE 11

Implicit Feedback and Ranking

  • Collabora0ve Filtering for Implicit Feedback Datasets
  • Y. Hu, Y. Koren, C. Volinsky
  • Implicit-to-Explicit Ordinal Logis0c Regression
  • D. Parra, A. Karatzoglou, X. Amatriain, I. Yavuz
  • BPR - Bayesian Personalized Ranking
  • S. Rendle, C. Freudenthaler, Z. Gantner, and L. S. Thieme
  • RankALS – Alterna0ng Least Squares for Personalized Ranking
  • G. Takacs, D. Tikk
  • CLiMF – Reciprocal Rank Op0miza0on

Y Shi, A. Karatzoglou, L. Baltrunas, M. Larson, N. Oliver, A. Hanjalic

11

slide-12
SLIDE 12

ONE-CLASS COLLABORATIVE FILTERING WITH RANDOM GRAPHS

12

Ulrich Paquet and Noam Koenigstein Interna'onal World Wide Web Conference (WWW'13), May 2013, Rio de Janeiro, Brazil.

slide-13
SLIDE 13

Problem FormulaMon

... N ≈ 10 – 100K nodes M ≈ 10 – 100M nodes

? ? ? ?

BiparMte graph → We care about ? = p(link)

slide-14
SLIDE 14

The Hidden Graph

𝐻={​𝑕↓𝑛𝑜 }, 𝐼={​ℎ↓𝑛𝑜 }

edges 𝑕,ℎ ∈{0,1} ... ​𝑕↓𝑛𝑜 =0

​ℎ↓𝑛𝑜 =1 ​𝑕↓𝑛𝑜 =0 ​ℎ↓𝑛𝑜 =0 ​𝐯↓𝑛 ​𝐰↓𝑜 𝑞​𝑕=1 ⁠ 𝐯,𝐰,ℎ=1 = 𝜏(​𝐯↑𝑈 𝐰)

1

​𝐯↑𝑈 𝐰 ​𝑕↓𝑛𝑜 =1 ​ℎ↓𝑛𝑜 =1

slide-15
SLIDE 15

BESIDES FEEDBACK: COLD START, META-DATA, HYBRID, CONTEXTUAL…

15

slide-16
SLIDE 16

XBOX MOVIES RECOMMENDATIONS: VARIATIONAL BAYES MATRIX FACTORIZATION WITH EMBEDDED FEATURE SELECTION

16

Noam Koenigstein and Ulrich Paquet ACM Conference on Recommender Systems (RecSys'13), October 2013, Hong Kong, China.

slide-17
SLIDE 17

Movie Features (tags)

17

Categories: § Plot § Mood § Audience § Time Period Harry Pocer and the Philosopher's Stone

  • Imaginary
  • Wizards and Magicians
  • Best Friends
  • Exci0ng
  • Humorous
  • Danger
  • Kids
  • Teens
  • Contemporary
  • 21st Century
slide-18
SLIDE 18

18

𝑞(​𝑔 ​𝑔↓1 ,​𝑔 ​𝑔↓2 |𝛽=0.01, =0.01,𝛾=0.01) =0.01)

slide-19
SLIDE 19

19

  • 10
  • 5

5 10 15

  • 30
  • 20
  • 10

10 20 30 40 50

Kids Semi Fantastic New Wave India Pets Adventure Foreign Rescue Drugs/Alcohol Semi Serious Animal life Profanity Serial Killer Scary Sweden Sexy Experimental B&W Erotic Suspenseful Family Gatherings Cannes Festival Winner Australia Grossout Humor Horror

slide-20
SLIDE 20

GROOVE RADIO: A BAYESIAN HIERARCHICAL MODEL FOR PERSONALIZED PLAYLIST GENERATION

20

Shay Ben-Elazar, Gal Lavee, Noam Koenigstein, Oren Barkan, Hilik Berezin, Ulrich Paquet, Tal Zaccai ACM Conference on Web Search and Data Mining (WSDM'17), Cambridge UK, February 2017.

slide-21
SLIDE 21

THE GAP BETWEEN COLLABORATIVE FILTERING RESEARCH AND REAL WORLD RECOMMENDATIONS

21

slide-22
SLIDE 22

The Gap Between CollaboraMve Filtering and Real Recommenders

  • Diversity vs. accuracy - tradeoff??
  • Popularity vs. personaliza0on
  • Item fa0gue / freshness – repea0ng items
  • Serendipity – when and how much to “surprise” the user
  • List Recommenda0ons / page op0miza0on
  • Predic0ng the future vs. influencing the user
  • Metrics and Evalua0on

22

slide-23
SLIDE 23

The Salesperson Analogy

23

slide-24
SLIDE 24

BEYOND COLLABORATIVE FILTERING: THE LIST RECOMMENDATION PROBLEM

24

Oren Sar Shalom, Noam Koenigstein, Ulrich Paquet, Hastagiri P. Vanchinathan Interna'onal World Wide Web Conference (WWW'16), April 2016, Montreal, Canada.

slide-25
SLIDE 25

List RecommendaMons in Xbox 360

25

slide-26
SLIDE 26

Conclusions

  • There is s0ll a gap between most CF models and the actual goal of recommender systems
  • Learning individual user-item tuples or ranking preferences is problema0c because:

– Can’t handle the diversity vs. accuracy “tradeoff” – List recommenda0ons / Page op0miza0on

  • Learning to predict future events from historical data is insufficient because:

– Can’t handle balancing popularity and personaliza0on – Freshness / Item Fa0gue – Serendipity

  • RL alone is not the ul0mate solu0on because:

– The abundance of implicit data – Represen0ng the “taste space”

  • Offline evalua0on metrics are insufficient

– They measure our ability to predict the future but not our ability to change it (influence the user)

  • Botom line: We s0ll have a lot to work in the RecSys community!

26

slide-27
SLIDE 27

Thank You!

27

We are looking for postdocs in Israel!!! Interested? Find me during the coffee break….