COSNET: Connecting Heterogeneous Social Networks with...

1

COSNET: Connecting Heterogeneous Social Networks with Local and Global Consistency

Yutao Zhang+, Jie Tang+, Zhilin Yang+, Jian Pei#, and Philip S. Yu*

+Tsinghua University #Simon Fraser University *University of Illinois at Chicago

2

p Academic Social Network Analysis and Mining system—AMiner (http://aminer.org) p  Online since 2006 p  >38 million researcher profiles p  >100 million publications p  >241 million requests p  >12.35 Terabyte data p  100K IP access from 170 countries

per month p  10% increase of visits per month

p Deep analysis, mining, and search

AMiner II (ArnetMiner)

3

Ruud Bolle Office: 1S-D58 Letters: IBM T.J. Watson Research Center P.O. Box 704 Yorktown Heights, NY 10598 USA Packages: IBM T.J. Watson Research Center 19 Skyline Drive Hawthorne, NY 10532 USA Email: [email protected] Ruud M. Bolle was born in Voorburg, The Netherlands. He received the Bachelor's Degree in Analog Electronics in 1977 and the Master's Degree in Electrical Engineering in 1980, both from Delft University of Technology, Delft, The Netherlands. In 1983 he received the Master's Degree in Applied Mathematics and in 1984 the Ph.D. in Electrical Engineering from Brown University, Providence, Rhode Island. In 1984 he became a Research Staff Member at the IBM Thomas J. Watson Research Center in the Artificial Intelligence Department of the Computer Science Department. In 1988 he became manager of the newly formed Exploratory Computer Vision Group which is part of the Math Sciences Department.

Currently, his research interests are focused on video database indexing, video processing, visual human-computer interaction and biometrics applications.

Ruud M. Bolle is a Fellow of the IEEE and the AIPR. He is Area Editor of Computer Vision and Image Understanding and Associate Editor of Pattern Recognition. Ruud M. Bolle is a Member of the IBM Academy of Technology.

DBLP: Ruud Bolle 2006

Nalini K. Ratha, Jonathan Connell, Ruud M. Bolle, Sharat Chikkerur: Cancelable Biometrics: A Case Study in Fingerprints. ICPR (4) 2006: 370-373EE50

Sharat Chikkerur, Sharath Pankanti, Alan Jea, Nalini K. Ratha, Ruud M. Bolle: Fingerprint Representation Using Localized Texture Features. ICPR (4) 2006: 521-524EE49

Andrew Senior, Arun Hampapur, Ying-li Tian, Lisa Brown, Sharath Pankanti, Ruud M. Bolle: Appearance models for occlusion handling. Image Vision Comput. 24(11): 1233-1243 (2006)EE48

2005Ruud M. Bolle, Jonathan H. Connell, Sharath Pankanti, Nalini K. Ratha, Andrew W. Senior: The Relation between the ROC Curve and the CMC. AutoID 2005: 15-20EE47

Sharat Chikkerur, Venu Govindaraju, Sharath Pankanti, Ruud M. Bolle, Nalini K. Ratha: Novel Approaches for Minutiae Verification in Fingerprint Images. WACV. 2005: 111-116EE46

...

Ruud Bolle Office: 1S-D58 Letters: IBM T.J. Watson Research Center P.O. Box 704 Yorktown Heights, NY 10598 USA Packages: IBM T.J. Watson Research Center 19 Skyline Drive Hawthorne, NY 10532 USA Email: [email protected] Ruud M. Bolle was born in Voorburg, The Netherlands. He received the Bachelor's Degree in Analog Electronics in 1977 and the Master's Degree in Electrical Engineering in 1980, both from Delft University of Technology, Delft, The Netherlands. In 1983 he received the Master's Degree in Applied Mathematics and in 1984 the Ph.D. in Electrical Engineering from Brown University, Providence, Rhode Island. In 1984 he became a Research Staff Member at the IBM Thomas J. Watson Research Center in the Artificial Intelligence Department of the Computer Science Department. In 1988 he became manager of the newly formed Exploratory Computer Vision Group which is part of the Math Sciences Department.

Currently, his research interests are focused on video database indexing, video processing, visual human-computer interaction and biometrics applications.

Ruud M. Bolle is a Fellow of the IEEE and the AIPR. He is Area Editor of Computer Vision and Image Understanding and Associate Editor of Pattern Recognition. Ruud M. Bolle is a Member of the IBM Academy of Technology.

Knowledge Acquisition from the Web (ACM TKDD, WWW’12, ISWC’06, ICDM’07, ACL’07)

Contact Information

Educational history

Academic services

Publications

1

1 2

2

Ruud Bolle

Position

Affiliation

Address

Address

Email

PhdunivPhdmajor

Phddate

Msuniv

Msdate

Msmajor

BsunivBsdate

Bsmajor

Research Staff

IBM T.J. Watson Research CenterP.O. Box 704Yorktown Heights, NY 10598 USA

[email protected]

Brown University

1984

Electrical Engineering

Delft University of Technology

Analog Electronics

1977

Delft University of Technology

IBM T.J. Watson Research Center19 Skyline DriveHawthorne, NY 10532 USA

IBM T.J. Watson Research Center

Electrical Engineering

1980

Applied Mathematics

Msmajor

http://researchweb.watson.ibm.com/ecvg/people/bolle.html

Homepage

Ruud BolleName

video database indexingvideo processing

visual human-computer interactionbiometrics applications

Research_Interest

Photo

Publication 1#

Cancelable Biometrics: A Case Study in Fingerprints

ICPR 370

2006

Date

Start_page

Venue

Title

373

End_page

Publication 2#

Fingerprint Representation Using Localized Texture Features

ICPR 521

2006

Date

Start_page

Venue

Title

524

End_page

. . .

Co-authorCo-author

1

Ruud Bolle

2 Publication #3

Publication #5

coauthor

coauthor

UIUC affiliation

Professor position

2

1

4

Researcher Profile Database[1]

Extracted more than 1,000,000 researcher profiles from the Web

[1] J. Tang, L. Yao, D. Zhang, and J. Zhang. A Combination Approach to Web User Profiling. ACM Transactions on Knowledge Discovery from Data (TKDD), (vol. 5 no. 1), Article 2 (December 2010), 44 pages.

5

Is this Enough?

6

Required semantics are distributed in multiple sources

LinkedIn Videolectures

7

Identity Linking •  Identifying users from multiple heterogeneous networks and integrating

semantics from the different networks together.

8

COSNET: Connecting Social Networks with Local and Global Consistency

•  Input: G={G1, G2, …, Gm}, with Gk=(Vk, Ek, Rk) •  Formalization: X={xi}, all possible pairwise

matchings and each corresponds to •  COSNET: an energy-based model

[1] Yutao Zhang, Jie Tang, Zhilin Yang, Jian Pei, and Philip Yu. COSNET: Connecting Heterogeneous Social Networks with Local and Global Consistency. KDD’15.

yi ∈{1,0}

Y * = argminE(Y ,X)

9

Local vs. Global consistency

•  Given three networks,

𝑣13

𝑣11

𝑣31

𝑣12

𝑣22

𝑣21

𝑣33 𝑣23 𝑣32

𝐺 2

Username: Ortiz_BrandyNation: USAGender: female

Username: @ortizbrandyNation: USAGender: female

𝐺 1

𝐺 3

10

Local vs. Global consistency •  Local matching: matching users by profiles

𝑣13

𝑣11

𝑣31

𝑣12

𝑣22

𝑣21

𝑣33 𝑣23 𝑣32

𝐺 2

局部一致性



𝐺 1

𝐺 3

Local consistency

Pairwise similarity features –  Username similarity and

uniqueness –  Profile content similarity –  Ego network similarity –  Social status

Energy function

11

𝑣13

𝑣11

𝑣31

𝑣12

𝑣22

𝑣21

𝑣33 𝑣23 𝑣32

𝐺 2

局部一致性网络一致性



𝐺 1

𝐺 3

Local vs. Global consistency •  Network matching: matching users’ ego networks

Network matching

Local consistency

Encourage “neighborhood-preserving matching”

12

𝑣13

𝑣11

𝑣31

𝑣12

𝑣22

𝑣21

𝑣33 𝑣23 𝑣32

𝐺 2




𝐺 1

𝐺 3


Network matching

Local consistency

𝑣11

𝑣12 𝑣22

𝑣21

True True

13

𝑣13

𝑣11

𝑣31

𝑣12

𝑣22

𝑣21

𝑣33 𝑣23 𝑣32

𝐺 2




𝐺 1

𝐺 3


Network matching

Local consistency

𝑣11

𝑣12 𝑣22

𝑣21

True True

𝑣11

𝑣12 𝑣22

𝑣21

False True x

14

𝑣13

𝑣11

𝑣31

𝑣12

𝑣22

𝑣21

𝑣33 𝑣23 𝑣32

𝐺 2




𝐺 1

𝐺 3


Network matching

Local consistency

𝑣11

𝑣12 𝑣22

𝑣21

True True

𝑣11

𝑣12 𝑣22

𝑣21

False True x

𝑣11

𝑣12 𝑣22

𝑣21

False False x x

15

𝑣13

𝑣11

𝑣31

𝑣12

𝑣22

𝑣21

𝑣33 𝑣23 𝑣32

𝐺 2




𝐺 1

𝐺 3


Network matching

Local consistency

𝑣11

𝑣12 𝑣22

𝑣21

True True

𝑣11

𝑣12 𝑣22

𝑣21

False True x

𝑣11

𝑣12 𝑣22

𝑣21

False False x x

16

Network Matching •  Network matching: matching users’ ego networks

𝒗𝟏𝟏

𝒗𝟏𝟐 𝒗𝟏𝟑

𝒗𝟐𝟏

𝒗𝟐𝟐 𝒗𝟐𝟑

𝒗𝟏𝟏𝒗𝟐𝟏 𝒗𝟏𝟐𝒗𝟐𝟏

𝒗𝟏𝟏𝒗𝟐𝟐 𝒗𝟏𝟑𝒗𝟐𝟑 𝒗𝟏𝟐𝒗𝟐𝟐

𝒗𝟏𝟐𝒗𝟐𝟑 𝒗𝟏𝟏𝒗𝟐𝟑

𝒗𝟏𝟑𝒗𝟐𝟏

𝒗𝟏𝟑𝒗𝟐𝟐

𝑮𝟏

𝑮𝟐

𝑪

Input networks Matching graph

Energy function

17

Candidate Pruning

•  Content-based method – Username similarity above a threshold

•  Structure-based similarity – Starting from a seed mapping set and iteratively

propagate the m

18

𝑣13

𝑣11

𝑣31

𝑣12

𝑣22

𝑣21

𝑣33 𝑣23 𝑣32

𝐺 2


全局一致性



𝐺 1

𝐺 3

Local vs. Global consistency •  Global consistency: matching users by avoiding global

inconsistency

Global inconsistency

Network matching

Local consistency

Avoid “global inconsistency”

19

𝑣13

𝑣11

𝑣31

𝑣12

𝑣22

𝑣21

𝑣33 𝑣23 𝑣32

𝐺 2


全局一致性



𝐺 1

𝐺 3


inconsistency


Network matching

Local consistency

𝑣13

𝑣11

𝑣12

𝑣13

𝑣11

𝑣12

𝑣13

𝑣11

𝑣12

𝑣13

𝑣11

𝑣12

x

x x

x x

x

20

𝑣13

𝑣11

𝑣31

𝑣12

𝑣22

𝑣21

𝑣33 𝑣23 𝑣32

𝐺 2


全局一致性



𝐺 1

𝐺 3


inconsistency


Network matching

Local consistency

𝑣13

𝑣11

𝑣12

𝑣13

𝑣11

𝑣12

𝑣13

𝑣11

𝑣12

𝑣13

𝑣11

𝑣12

x

x x

x x

x

Inconsistent!

21

Avoid Global Inconsistency

Energy function Input networks Matching graph

22

Model Construction

Objective function by combining all the energy functions

𝑣31

𝑣12

𝑣22

𝑣21

𝑣32

𝐺 2

s

𝑣11𝑣12

𝑣11

Matching Graph Generation

Candidate Pruning

Model Construction

𝑣11𝑣22 𝑣11𝑣32

𝑣21𝑣22 𝑣21𝑣12 𝑣21𝑣32

𝑣31𝑣12 𝑣31𝑣22 𝑣31𝑣32 𝑥1 𝑥2 𝑥3

𝑥4 𝑥5

𝑦1

𝑦4 𝑦5

𝑦2 𝑦3

𝑓 𝑙 (𝑥1 ) 𝑓 𝑙 (𝑥2 ) 𝑓 𝑙 (𝑥4 ) 𝑓 𝑙 (𝑥5 )

𝑓 𝑙 (𝑥3 )

𝑓 𝑒 (𝑦1 , 𝑦2 )

𝐺 1

𝑣11𝑣12 𝑣11𝑣32

𝑣21𝑣22

𝑣31𝑣12 𝑣31𝑣32

𝑣11𝑣12 𝑣11𝑣32 𝑣21𝑣22

𝑣31𝑣12 𝑣31𝑣32

𝑓 𝑒 (𝑦2 , 𝑦3 )

𝑓 𝑒 (𝑦2 , 𝑦4 ) 𝑓 𝑒 (𝑦2 , 𝑦5 )

(a) Two input networks (b) The generated matching graph (c) Matching graph after pruning (d) The constructed model

23

Model Learning •  Max-margin learning

•  As the original problem is intractable, we use Lagrangian relaxation to decompose the original objective function into a set of easy-to-solve sub-problems

24

Model Learning (cont.)

•  Dual decomposition

The resulting objective function is convex and non-differentiable, and can be solved by projected sub-gradient method

This provides a lower bound to the original function

25

Results

26

Connecting AMiner with …

•  LinkedIn and VideoLectures

Name-match: match name only; SVM: use classifier to identify the same user; MNA: an optimization method;

SiGMa: local propagation; COSNET: our method; COSNET-: w/o global consistency.

27

Connecting Social Media Sites

•  Twitter, LiveJournal, Last.fm, Flickr, MySpace

Name-match: match name only; SVM: use classifier to identify the same user; MNA: an optimization method;

SiGMa: local propagation; COSNET: our method; COSNET-: w/o global consistency.

28

Effects of Global Consistency

+5.4%

Academia Collection SNS Collection

+9.5%

COSNET-: w/o global consistency.

29

Application in AMiner

- Video contents

- Personal profiles- Business connections- Skills and expertise

- Patents data

30

Thanks!

http://aminer.org

Data & source code http://aminer.org/cosnet

COSNET: Connecting Heterogeneous Social Networks with...

Documents

Transcript of COSNET: Connecting Heterogeneous Social Networks with...