Deep Learning for Recommender Systems Justin Basilico & Yves - - PowerPoint PPT Presentation

▶
deep learning for recommender systems
SMART_READER_LITE
LIVE PREVIEW

Deep Learning for Recommender Systems Justin Basilico & Yves - - PowerPoint PPT Presentation

Aug 06, 2023 116 likes •481 views

Deep Learning for Recommender Systems Justin Basilico & Yves Raimond March 28, 2018 GPU Technology Conference @JustinBasilico @moustaki The value of recommendations A few seconds to find something great to watch Can only show a

slide-1
SLIDE 1

Deep Learning for Recommender Systems

Justin Basilico & Yves Raimond

March 28, 2018

GPU Technology Conference @JustinBasilico @moustaki

slide-2
SLIDE 2

The value of recommendations

  • A few seconds to find something

great to watch…

  • Can only show a few titles
  • Enjoyment directly impacts

customer satisfaction

  • Generates over $1B per year of

Netflix revenue

  • How? Personalize everything
slide-3
SLIDE 3

Deep learning for recommendations: a first try

slide-4
SLIDE 4

1 1 1 1 1 1 1 1 1 Users Items

Traditional Recommendation Setup

slide-5
SLIDE 5

U

≈

R V

A Matrix Factorization view

slide-6
SLIDE 6

U

A Feed-Forward Network view

V

slide-7
SLIDE 7

U

A (deeper) feed-forward view

V

Mean squared loss

?

slide-8
SLIDE 8

A quick & dirty experiment

  • ○

○

  • ○

■ ■ ○ ■ ■ ■ ■ ■

slide-9
SLIDE 9

GPU vs. CPU

slide-10
SLIDE 10
slide-11
SLIDE 11
slide-12
SLIDE 12

What’s going on?

slide-13
SLIDE 13

Conclusion?

slide-14
SLIDE 14

Breaking the ‘traditional’ recsys setup

slide-15
SLIDE 15

Alternative data

slide-16
SLIDE 16

Content-based side information

slide-17
SLIDE 17

Metadata-based side information

  • ○
  • ○
slide-18
SLIDE 18

YouTube Recommendations

slide-19
SLIDE 19

Alternative models

slide-20
SLIDE 20

Restricted Boltzmann Machines

slide-21
SLIDE 21

Auto-encoders

  • ○
slide-22
SLIDE 22

(*)2Vec

  • prod2vec

(Skip-gram)

user2vec

(Continuous Bag of Words)

slide-23
SLIDE 23

Wide + Deep models

  • [Cheng et. al., 2016]
slide-24
SLIDE 24

Alternative framings

slide-25
SLIDE 25

Sequence prediction

  • ○

○

  • ○

○

slide-26
SLIDE 26

Contextual sequence prediction

slide-27
SLIDE 27

Contextual sequence data

2017-12-10 15:40:22 2017-12-23 19:32:10 2017-12-24 12:05:53 2017-12-27 22:40:22 2017-12-29 19:39:36 2017-12-30 20:42:13 Context Action Sequence per user

?

Time

slide-28
SLIDE 28

Time-sensitive sequence prediction

  • ○
  • ○

■

  • ■

○

slide-29
SLIDE 29
slide-30
SLIDE 30
slide-31
SLIDE 31

Other framings

  • ○
  • ○
slide-32
SLIDE 32

Conclusion

slide-33
SLIDE 33

Takeaways

slide-34
SLIDE 34

More Resources

slide-35
SLIDE 35

Thank you.

@JustinBasilico @moustaki

Justin Basilico & Yves Raimond

Yes, we’re hiring...