84
Groups and Reputation in Social Networks V.A. Traag Promoteurs: Paul Van Dooren, Yurii Nesterov Jury: François Glineur Vincent Blondel Marco Saerens Patrick De Leenheer ICTEAM Université Catholique de Louvain 2 September 2013 Soutenance Publique

Public thesis defence: groups and reputation in social networks

Embed Size (px)

DESCRIPTION

Public defence of my thesis on groups and reputation in social networks (2 september 2013, Louvain-la-Neuve).

Citation preview

Page 1: Public thesis defence: groups and reputation in social networks

Groups and Reputation in Social Networks

V.A. Traag

Promoteurs:Paul Van Dooren, Yurii Nesterov

Jury:François GlineurVincent BlondelMarco Saerens

Patrick De Leenheer

ICTEAMUniversité Catholique de Louvain

2 September 2013

Soutenance Publique

Page 2: Public thesis defence: groups and reputation in social networks

Overview

Social Networks

1 2Community Detection Reputation & Negative Links

Page 3: Public thesis defence: groups and reputation in social networks

Overview

Social Networks

1 2Community Detection Reputation & Negative Links

Page 4: Public thesis defence: groups and reputation in social networks

Social networks

Page 5: Public thesis defence: groups and reputation in social networks

Social networks

Page 6: Public thesis defence: groups and reputation in social networks

Social networks

Page 7: Public thesis defence: groups and reputation in social networks

Social networks

Page 8: Public thesis defence: groups and reputation in social networks

Social networks

Page 9: Public thesis defence: groups and reputation in social networks

Social networks

Page 10: Public thesis defence: groups and reputation in social networks

Social networks

Page 11: Public thesis defence: groups and reputation in social networks

Social networks

Page 12: Public thesis defence: groups and reputation in social networks

Social networks

Page 13: Public thesis defence: groups and reputation in social networks

Social networks

Page 14: Public thesis defence: groups and reputation in social networks

Social networks

I did not analyse these

Page 15: Public thesis defence: groups and reputation in social networks

Keep it simple

Page 16: Public thesis defence: groups and reputation in social networks

Keep it simple

Page 17: Public thesis defence: groups and reputation in social networks

Keep it simple

Page 18: Public thesis defence: groups and reputation in social networks

Keep it simple

Page 19: Public thesis defence: groups and reputation in social networks

Keep it simple

Node or Vertex

Page 20: Public thesis defence: groups and reputation in social networks

Keep it simple

Link or Edge

Page 21: Public thesis defence: groups and reputation in social networks

Keep it simple

Network or Graph

Page 22: Public thesis defence: groups and reputation in social networks

Overview

Social Networks

1 2Community Detection Reputation & Negative Links

Page 23: Public thesis defence: groups and reputation in social networks

Overview

Social Networks

1 2Community Detection Reputation & Negative Links

Page 24: Public thesis defence: groups and reputation in social networks

What is a community?

0

12

3

4

56

7

8

9

10

11

Community or Cluster

Page 25: Public thesis defence: groups and reputation in social networks

What is a community?

0

12

3

4

56

7

8

9

10

11

Community or Cluster

But is it a good community?

Page 26: Public thesis defence: groups and reputation in social networks

What is a community?

0

12

3

4

56

7

8

9

10

11

Community or Cluster

But is it a good community?

Count links in community

Page 27: Public thesis defence: groups and reputation in social networks

What is a community?

0

12

3

4

56

7

8

9

10

11

Community or Cluster

But is it a good community?

Count links in communitymc = 4

Page 28: Public thesis defence: groups and reputation in social networks

What is a community?

0

12

3

4

56

7

8

9

10

11

Community or Cluster

But is it a good community?

Count links in communitymc = 4

Possible links?

Page 29: Public thesis defence: groups and reputation in social networks

What is a community?

0

12

3

4

56

7

8

9

10

11

Community or Cluster

But is it a good community?

Count links in communitymc = 4

Possible links?(nc2

)= nc(nc−1)

2 = 4×32 = 6

Page 30: Public thesis defence: groups and reputation in social networks

What is a community?

0

12

3

4

56

7

8

9

10

11

Community or Cluster

But is it a good community?

Count links in communitymc = 4

Possible links?(nc2

)= nc(nc−1)

2 = 4×32 = 6

4 links in community,2 missing.

Page 31: Public thesis defence: groups and reputation in social networks

What is a community?

0

12

3

4

56

7

8

9

10

11

Community or Cluster

But is it a good community?

Count links in communitymc = 4

Possible links?(nc2

)= nc(nc−1)

2 = 4×32 = 6

4 links in community,2 missing.

Want many links, few missing

Page 32: Public thesis defence: groups and reputation in social networks

What is a community?

0

12

3

4

56

7

8

9

10

11

Bad community

Count links in communitymc = 5

Possible links?(nc2

)= nc(nc−1)

2 = 5×42 = 10

5 links in community,5 missing.

Page 33: Public thesis defence: groups and reputation in social networks

What is a community?

0

12

3

4

56

7

8

9

10

11

Are there better communities?

Page 34: Public thesis defence: groups and reputation in social networks

Make money

0

12

3

4

56

7

8

9

10

11

Get money for communities:

• get 1 − c for every edge in community,

• pay c for missing edge in community.

Page 35: Public thesis defence: groups and reputation in social networks

Make money

0

12

3

4

56

7

8

9

10

11

Get money for communities:

• get 1 − c for every edge in community,

• pay c for missing edge in community.

Suppose c = e0.50

4 edges, 2 missing:4 × 0.5 − 2 × 0.5 = e1.00

Page 36: Public thesis defence: groups and reputation in social networks

Make money

0

12

3

4

56

7

8

9

10

11

Get money for communities:

• get 1 − c for every edge in community,

• pay c for missing edge in community.

Suppose c = e0.50

4 edges, 2 missing:4 × 0.5 − 2 × 0.5 = e1.00

2 edges, 1 missing:2 × 0.5 − 1 × 0.5 = e0.50

Page 37: Public thesis defence: groups and reputation in social networks

Make money

0

12

3

4

56

7

8

9

10

11

Get money for communities:

• get 1 − c for every edge in community,

• pay c for missing edge in community.

Suppose c = e0.50

4 edges, 2 missing:4 × 0.5 − 2 × 0.5 = e1.00

2 edges, 1 missing:2 × 0.5 − 1 × 0.5 = e0.50

5 edges, 5 missing:5 × 0.5 − 5 × 0.5 = e0.00

Page 38: Public thesis defence: groups and reputation in social networks

Make money

0

12

3

4

56

7

8

9

10

11

Get money for communities:

• get 1 − c for every edge in community,

• pay c for missing edge in community.

Suppose c = e0.50

4 edges, 2 missing:4 × 0.5 − 2 × 0.5 = e1.00

2 edges, 1 missing:2 × 0.5 − 1 × 0.5 = e0.50

5 edges, 5 missing:5 × 0.5 − 5 × 0.5 = e0.00

Total: e1.50

Page 39: Public thesis defence: groups and reputation in social networks

Better communities

Initial communities: e1.50

0

12

3

4

56

7

8

9

10

11

Page 40: Public thesis defence: groups and reputation in social networks

Better communities

Initial communities: e1.50

Move 0: get 2 × 0.5 = e1.00 extra

0

12

3

4

56

7

8

9

10

11

Page 41: Public thesis defence: groups and reputation in social networks

Better communities

Initial communities: e1.50

Move 0: get 2 × 0.5 = e1.00 extra

Move 5: get 3 × 0.5 = e1.50 extra

0

12

3

4

56

7

8

9

10

11

Page 42: Public thesis defence: groups and reputation in social networks

Better communities

Initial communities: e1.50

Move 0: get 2 × 0.5 = e1.00 extra

Move 5: get 3 × 0.5 = e1.50 extra

Move 11: get 1 × 0.5 = e0.50 extra

0

12

3

4

56

7

8

9

10

11

Page 43: Public thesis defence: groups and reputation in social networks

Better communities

Initial communities: e1.50

Move 0: get 2 × 0.5 = e1.00 extra

Move 5: get 3 × 0.5 = e1.50 extra

Move 11: get 1 × 0.5 = e0.50 extra

0

12

3

4

56

7

8

9

10

11

Final communities: e4.50

Page 44: Public thesis defence: groups and reputation in social networks

Better communities

Initial communities: e1.50

Move 0: get 2 × 0.5 = e1.00 extra

Move 5: get 3 × 0.5 = e1.50 extra

Move 11: get 1 × 0.5 = e0.50 extra

0

12

3

4

56

7

8

9

10

11

Louvain Algorithm

Page 45: Public thesis defence: groups and reputation in social networks

Real network

My own Facebook

Page 46: Public thesis defence: groups and reputation in social networks

Real network

My own Facebook c = e0.–528 internal links

Page 47: Public thesis defence: groups and reputation in social networks

Real network

My own Facebook c = e0.0002527 internal links

Page 48: Public thesis defence: groups and reputation in social networks

Real network

My own Facebook c = e0.0025526 internal links

Page 49: Public thesis defence: groups and reputation in social networks

Real network

My own Facebook c = e0.01511 internal links

Page 50: Public thesis defence: groups and reputation in social networks

Real network

My own Facebook c = e0.10491 internal links

Page 51: Public thesis defence: groups and reputation in social networks

Real network

My own Facebook c = e0.500 internal links

Page 52: Public thesis defence: groups and reputation in social networks

Resolution limit

Page 53: Public thesis defence: groups and reputation in social networks

Significance

Original partition (14 links)

Page 54: Public thesis defence: groups and reputation in social networks

Significance

Original partition (14 links)

Random graph (5 links)

Page 55: Public thesis defence: groups and reputation in social networks

Significance

Original partition (14 links)

Random graph (5 links) Find partition (11 links)

Page 56: Public thesis defence: groups and reputation in social networks

Overview

Social Networks

1 2Community Detection Reputation & Negative Links

Page 57: Public thesis defence: groups and reputation in social networks

Overview

Social Networks

1 2Community Detection Reputation & Negative Links

Page 58: Public thesis defence: groups and reputation in social networks

Social Balance

+Leonard Sheldon

Page 59: Public thesis defence: groups and reputation in social networks

Social Balance

+

+

Leonard Sheldon

Leslie

Page 60: Public thesis defence: groups and reputation in social networks

Social Balance

+

+ −

Leonard Sheldon

Leslie

Page 61: Public thesis defence: groups and reputation in social networks

Social Balance

+

−−

Leonard Sheldon

Leslie

Page 62: Public thesis defence: groups and reputation in social networks

Social Balance

+

−−

Leonard Sheldon

LeslieThe enemy of my enemy is my friend

Page 63: Public thesis defence: groups and reputation in social networks

Social Balance

+

−−

Leonard Sheldon

LeslieThe friend of my enemy is my enemy

Page 64: Public thesis defence: groups and reputation in social networks

Social Balance

+

−−

Leonard Sheldon

LeslieThe friend of my enemy is my enemy

Socially balanced: +−− = +

Page 65: Public thesis defence: groups and reputation in social networks

Balanced Triads

Balanced Triads

Page 66: Public thesis defence: groups and reputation in social networks

Balanced Triads

Balanced Triads Unbalanced Triads

Page 67: Public thesis defence: groups and reputation in social networks

Balanced Triads

Balanced TriadsTwo Factions

Page 68: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−

Leonard Sheldon

Leslie

Page 69: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−+

Leonard Sheldon

Leslie

Page 70: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−+

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Page 71: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−+

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Ask Leonard what to do

Page 72: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−+

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Ask Leonard what to do

Sheldon likes Leonard +

Page 73: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−+

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Ask Leonard what to do

Sheldon likes Leonard +

Leonard likes Leslie +

Page 74: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−+

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Ask Leonard what to do

Sheldon likes Leonard +

Leonard likes Leslie +

Effect ++ = +

Page 75: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−+

−+

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Ask Leonard what to do

Sheldon likes Leonard +

Leonard likes Leslie +

Effect ++ = +

t

Page 76: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−+

−+

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Ask Leonard what to do

Sheldon likes Leonard +

Leonard likes Leslie +

Effect ++ = +

t

Page 77: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−+

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Ask Leonard what to do

Sheldon likes Leonard +

Leonard likes Leslie +

Effect ++ = +

Compare friends with Leslie

Page 78: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−+

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Ask Leonard what to do

Sheldon likes Leonard +

Leonard likes Leslie +

Effect ++ = +

Compare friends with Leslie

Sheldon likes Leonard +

Page 79: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−+

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Ask Leonard what to do

Sheldon likes Leonard +

Leonard likes Leslie +

Effect ++ = +

Compare friends with Leslie

Sheldon likes Leonard +

Leslie dislikes Leonard −

Page 80: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−+

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Ask Leonard what to do

Sheldon likes Leonard +

Leonard likes Leslie +

Effect ++ = +

Compare friends with Leslie

Sheldon likes Leonard +

Leslie dislikes Leonard −Effect +− = −

Page 81: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−+

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Compare friends with Leslie

Sheldon likes Leonard +

Leslie dislikes Leonard −Effect +− = −

t

Page 82: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Compare friends with Leslie

Sheldon likes Leonard +

Leslie dislikes Leonard −Effect +− = −

t

Page 83: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Compare friends with Leslie

Sheldon likes Leonard +

Leslie dislikes Leonard −Effect +− = −

t

Page 84: Public thesis defence: groups and reputation in social networks

Changing Triads

+

−−

Leonard Sheldon

LeslieSheldon’s relationship with Leslie changes

Ask Leonard what to do

Compare friends with Leslie

Does not converge tosocial balance

Does converge tosocial balance