Field-weighted Factorization Machines for Click-Through Rate - - PowerPoint PPT Presentation

▶
field weighted factorization machines for click through
SMART_READER_LITE
LIVE PREVIEW

Field-weighted Factorization Machines for Click-Through Rate - - PowerPoint PPT Presentation

Nov 19, 2023 23 likes •268 views

Field-weighted Factorization Machines for Click-Through Rate Prediction in Display Advertising Author : Junwei Pan, Jian Xu, Alfonso Lobos Ruiz, Wenliang Zhao, Shengjun Pan, Yu Sun, Quan Lu Source : WWW 18 Advisor : Jia-Ling Koh

slide-1
SLIDE 1

Field-weighted Factorization Machines for Click-Through Rate Prediction in Display Advertising

Author : Junwei Pan, Jian Xu, Alfonso Lobos Ruiz, Wenliang Zhao, Shengjun Pan, Yu Sun, Quan Lu Source : WWW’ 18 Advisor : Jia-Ling Koh Speaker : Chia-Yi Huang Date : 2018/05/22

slide-2
SLIDE 2

Outline

▸ Introduction

▸ Method ▸ Experiment ▸ Conclusion

2

slide-3
SLIDE 3

▸ Goal

  • Display ad CTR prediction use Field-

weighted Factorization Machines

Introduction

3

slide-4
SLIDE 4

▸ Goal

  • Display ad CTR prediction use Field-

weighted Factorization Machines

Introduction

4

slide-5
SLIDE 5

▸ Field & Feature

Introduction

5

slide-6
SLIDE 6

▸ Feature interactions are prevalent and need

to be specifically modeled.

▸ Features from one field often interact

differently with features from different other fields.

▸ Potentially high model complexity needs to

be taken care of.


Challenge

6

slide-7
SLIDE 7

▸ Factorization Machine(FM, 因子分解機) ▸ Field-aware Factorization

Machine(FFM)

▸ Field-weighted Factorization

Machine(FwFM)

Background

7

slide-8
SLIDE 8

Outline

▸ Introduction

▸ Method

▸ Experiment ▸ Conclusion

8

slide-9
SLIDE 9

9

Evolution

▸ Logistic Regression model ▸ Degree-2 Polynomial model

slide-10
SLIDE 10

10

Evolution

▸ Factorization Machine ▸ Field-aware Factorization

Machine(FFM)

slide-11
SLIDE 11

11

Evolution

▸ Field-weighted Factorization

Machine(FwFM)

slide-12
SLIDE 12

12

Mutual Information

slide-13
SLIDE 13

Outline

▸ Introduction ▸ Method

▸ Experiment

▸ Conclusion

13

slide-14
SLIDE 14

▸ Data sets

14

Experiment

slide-15
SLIDE 15

15

Experiment

▸ Comparison of FwFMs with Existing

Models.

slide-16
SLIDE 16

16

Experiment

▸ Comparison of FwFMs and FFMs

using the same number of parameters.

slide-17
SLIDE 17

▸ L2 Regularization


17

Experiment

slide-18
SLIDE 18

▸ Learning Rate

18

Experiment

slide-19
SLIDE 19

▸ Embedding Vector Dimension

19

Experiment

slide-20
SLIDE 20

▸ Learned field interaction strengths

  • For
  • For
  • For

20

Experiment

slide-21
SLIDE 21

21

Experiment

P1356

slide-22
SLIDE 22

22

Experiment

slide-23
SLIDE 23

Outline

▸ Introduction ▸ Method ▸ Experiment

▸ Conclusion

23

slide-24
SLIDE 24

Conclusion

▸ FwFMs are competitive to FFMs with

significantly less parameters.

▸ FwFMs can indeed learn different

feature interaction strengths from different field pairs.

24