Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele...

218
Network Analysis Graphs as Matrices Michele Coscia & Luca Rossi ITU København

Transcript of Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele...

Page 1: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Network Analysis

Graphs as Matrices

Michele Coscia & Luca RossiITU København

Page 2: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem
Page 3: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Background:

Page 4: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Background:

Looking at a problem from a different

perspective

Page 5: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Sergey Brin & Larry Page

(1973-)

Page 6: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem
Page 7: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem
Page 8: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

55 Million!

Page 9: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

55 Million!

Page 10: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

PageRank

Page 11: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

PageRank● p of visiting v…

Page 12: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

PageRank● p of visiting v…● … = p of visiting a neighbor u of v *

p of going from u to v

Page 13: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

PageRank● p of visiting v…● … = p of visiting a neighbor u of v *

p of going from u to v● for v in V…

Page 14: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

PageRank● p of visiting v…● … = p of visiting a neighbor u of v *

p of going from u to v● for v in V…● 55 million times!

Page 15: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

PageRankG = (V, E)

Page 16: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

PageRankG = (V, E) A = 1 if uv in E

Page 17: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

PageRankG = (V, E) A = 1 if uv in E

Page 18: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

PageRankG = (V, E) A = 1 if uv in E

Page 19: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

PageRank

for v in V...

G = (V, E) A = 1 if uv in E

Page 20: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

PageRank

for v in V... R := AR

G = (V, E) A = 1 if uv in E

~

Page 21: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

PageRank

for v in V... R := AR

G = (V, E) A = 1 if uv in E

~

Row-normalized A

Page 22: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

PageRank

for v in V... R := AR

G = (V, E) A = 1 if uv in E

~

Row-normalized A

Eigenvectorof A

Page 23: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Random Walks

Page 24: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Basics

A B C D E FA

B

C

D

E

F

A B

Page 25: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Basics

A B C D E FA

B

C

D

E

F

Do youknow each

other?

A B

Page 26: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Basics

A B C D E FA

B

C

D

E

F

Do youknow each

other?

A B

Yes! Yes!

Page 27: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Basics

A B C D E FA

B

C

D

E

F

Do youknow each

other?

A B

Yes! Yes!

X

X

Page 28: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Basics

Page 29: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Basics

1 2 3 4 5 6 7 8 9

Page 30: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Basics

1 2 3 4 5 6 7 8 9123456789

Page 31: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Basics

1 2 3 4 5 6 7 8 9123456789

Page 32: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Basics

1 2 3 4 5 6 7 8 9123456789

Page 33: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Basics

1 2 3 4 5 6 7 8 9123456789

Page 34: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Basics

1 2 3 4 5 6 7 8 9123456789

Page 35: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Basics

1 2 3 4 5 6 7 8 9123456789

Adjacency Matrix

Page 36: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Properties of Adjacency Matrices

Page 37: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Properties of Adjacency Matrices

3

Page 38: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Properties of Adjacency Matrices

3

3

2

1

Page 39: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Properties of Adjacency Matrices

Page 40: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stochastic Adjacency

A B C D E FA

B

C

D

E

F

AX X X

Page 41: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stochastic Adjacency

A B C D E FA

B

C

D

E

F

Pick oneof your friends

at random

AX X X

Page 42: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stochastic Adjacency

A B C D E FA

B

C

D

E

F

Pick oneof your friends

at random

A

Mmm…E!

X X X

Page 43: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stochastic Adjacency

Page 44: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stochastic Adjacency● p of going from

v to u

Page 45: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stochastic Adjacency● p of going from

v to u

● A / A.sum(axis = 1)

Page 46: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stochastic Adjacency● p of going from

v to u

● A / A.sum(axis = 1)

Page 47: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stochastic Adjacency● p of going from v

to u

● A / A.sum(axis = 1)

● Useful eigen-properties

Page 48: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk Probabilities

Page 49: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk Probabilities

Page 50: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk Probabilities

Page 51: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk Probabilities

Page 52: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk Probabilities

Page 53: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk probabilities

Page 54: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk probabilities

Page 55: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk probabilities

Page 56: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk probabilities

Page 57: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk probabilities

Page 58: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk probabilities

Page 59: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk probabilities

Page 60: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk probabilities

Page 61: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk probabilities

Page 62: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk probabilities

p(1→2) in two steps?

Page 63: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk probabilities

p(1→2) in two steps?

(p(1→5) * p(5→2)) + (p(1→9) * p(9→2))

Page 64: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Walk probabilities

p(1→2) in two steps?

(p(1→5) * p(5→2)) + (p(1→9) * p(9→2))

p(u→v)?

Page 65: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

A

Page 66: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

A

Page 67: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

A

Page 68: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

A A2

Page 69: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

A A2

Page 70: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

A A2

Page 71: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

A A2

Page 72: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

A A2

Page 73: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

Page 74: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

Page 75: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

Page 76: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

Page 77: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

1

Page 78: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

Page 79: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

Page 80: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Powers of Stochastic

Page 81: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

A2

Page 82: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

A2

Page 83: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

A2 A3

Page 84: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

A2 A30

Page 85: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution

Page 86: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution● A∞ = π

Page 87: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution● A∞ = π

Page 88: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution● A∞ = π

● πA = π

Page 89: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution● A∞ = π

● πA = π

● ~ PageRank

Page 90: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Disconnected Mazes

Page 91: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Disconnected Mazes

Page 92: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Disconnected Mazes

Page 93: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Disconnected Mazes

?

Page 94: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Disconnected Mazes

?

Page 95: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Multiple Components

A

Page 96: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Multiple Components

A30

Page 97: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Multiple Components

A30

Page 98: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Multiple Components

A30

Page 99: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Multiple Components

A30

Page 100: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Multiple Components

A30

Page 101: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Google’s Solution

Page 102: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Google’s Solution

Page 103: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Google’s Solution

Beamme up,Scotty!

Page 104: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Google’s Solution

Page 105: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Google’s Solution

A

Page 106: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Google’s Solution

A

Page 107: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Google’s Solution

A+ε

Page 108: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution vs Degree

Page 109: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution vs Degree

Page 110: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution vs Degree

Corr = 1

Page 111: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution vs Degree

Corr = 1π = k / k.sum()

Page 112: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution vs PageRank

Page 113: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution vs PageRank

Page 114: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution vs PageRank

Page 115: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution vs PageRank

Page 116: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Stationary Distribution vs PageRank

Page 117: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Graph Spectrum

Page 118: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

A Recap on Eigenvectors

v

(x, y)

Page 119: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

A Recap on Eigenvectors

v

w = Av

(x’, y’)

(x, y)

Page 120: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

A Recap on Eigenvectors

v

w = Av

(x’, y’)(λx, λy)

(x, y)

Page 121: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

A Recap on Eigenvectors

v

w = Av

(x’, y’)

w = Av = λv

(λx, λy)

(x, y)

Page 122: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

A Recap on Eigenvectors

v

w = Av

(x’, y’)

w = Av = λv

(λx, λy)

Eigenvector

(x, y)

Page 123: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

A Recap on Eigenvectors

v

w = Av

(x’, y’)

w = Av = λv

(λx, λy)

Eigenvector

Eigenvalue

(x, y)

Page 124: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Right & Left

Av = λv

Page 125: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Right & Left

Av = λv Right Eigenvector

Page 126: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Right & Left

Av = λv Right Eigenvector

Page 127: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Right & Left

Av = λv

vA = vλ

Right Eigenvector

Page 128: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Right & Left

Av = λv

vA = vλ

Right Eigenvector

Left Eigenvector

Page 129: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Right & Left

Av = λv

vA = vλ

Right Eigenvector

Left Eigenvector

Page 130: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Right & Left

Av = λv

vA = vλ

Right Eigenvector

Left Eigenvector

Page 131: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Right & Left

Av = λv

vA = vλ

Right Eigenvector

Left Eigenvector

Page 132: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

Page 133: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

1

Page 134: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

1

Page 135: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

Page 136: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

Page 137: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

Page 138: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

Page 139: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

Page 140: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

Page 141: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

Page 142: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

Page 143: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

If both equal to 1

Page 144: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

If both equal to 1Graph has two components

Page 145: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

If both equal to 1Graph has two components

Page 146: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

If both equal to 1Graph has two components

λ1

Page 147: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

If both equal to 1Graph has two components

λ1

λ2

Page 148: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

λ1 > λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

Page 149: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

λ1 > λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

Page 150: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues & Components

λ1 > λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

Page 151: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

Page 152: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

v1 v

2 v

3 … v

n-2 v

n-1 v

n

Page 153: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

v1 v

2 v

3 … v

n-2 v

n-1 v

n

“Largest” Eigenvector

Page 154: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

vi

a

b

c

d

e

f

g

h

i

l

m

n

o

Page 155: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

vi

a

b

c

d

e

f

g

h

i

l

m

n

o

Page 156: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

vi

a

b

c

d

e

f

g

h

i

l

m

n

o

Page 157: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

vi

a

b

c

d

e

f

g

h

i

l

m

n

o

Page 158: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

vi

a

b

c

d

e

f

g

h

i

l

m

n

o

Page 159: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors & Components

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

If both equal to 1Graph has two components

λ1

λ2

Page 160: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors & Components

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

If both equal to 1Graph has two components

λ1

λ2

v1 v

2

Page 161: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors & Components

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

If both equal to 1Graph has two components

λ1

λ2

v1 v

2

a

a

a

a

a

a

0

0

0

0

0

0

0

0

0

b

b

b

Page 162: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors & Components

λ1 > λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

Page 163: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors & Components

λ1 > λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

v1

Page 164: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors & Components

λ1 > λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

v1

λ1

Page 165: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors & Components

λ1 > λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

v1

a

a

a

a

a

a

a

a

a

λ1

Page 166: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left Largest Eigenvector

Page 167: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left Largest Eigenvector

Page 168: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left Largest Eigenvector

Page 169: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left Largest Eigenvector

vL1 = π

Page 170: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left Largest Eigenvector

vL1 = π

πA = π

Page 171: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left Largest Eigenvector

vL1 = π

πA = π

πA = π1

Page 172: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left Largest Eigenvector

vL1 = π

πA = π

πA = π1

λ1

Page 173: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left Largest Eigenvector

vL1 = π

πA = π

πA = π1

λ1

Page 174: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left & Right

Page 175: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left & RightvL

1

Page 176: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left & RightvL

1

Page 177: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left & RightvL

1

v1

a

a

a

a

a

a

a

a

a

Page 178: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left & RightvL

1

vR1

v1

a

a

a

a

a

a

a

a

a

Page 179: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left & RightvL

1

vR1 A

Page 180: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left & RightvL

1

vR1 AT

Page 181: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left & RightvL

1

vR1 AT

A / A.sum(axis = 0)

Page 182: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left & Right

AT

A / A.sum(axis = 0)

Page 183: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Left & Right

AT

A / A.sum(axis = 0)

vL1

vR1

Page 184: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Graph Laplacian

Page 185: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Degree Matrix

Page 186: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Degree MatrixZeros everywhere

Page 187: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Degree MatrixZeros everywhere

i-th diagonal element...

Page 188: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Degree MatrixZeros everywhere

i-th diagonal element... … is the degree of node i

Page 189: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Degree MatrixZeros everywhere

i-th diagonal element... … is the degree of node i

D

Page 190: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Laplacian Matrix

D A

Page 191: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Laplacian Matrix

D AL = -

Page 192: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Laplacian Matrix

L = D - A

Page 193: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Laplacian Matrix

L = D - A

Page 194: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Laplacian Matrix

L = D - A

Page 195: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Laplacian Matrix

L = D - A

0

Page 196: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

1

A

L

Page 197: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

1

A

L λ1 ≤ λ

2 ≤ λ

3 ≤ … ≤ λ

n-2 ≤ λ

n-1 ≤ λ

n

Page 198: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

1

A

L λ1 ≤ λ

2 ≤ λ

3 ≤ … ≤ λ

n-2 ≤ λ

n-1 ≤ λ

n

0

Page 199: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvalues

λ1 ≥ λ

2 ≥ λ

3 ≥ … ≥ λ

n-2 ≥ λ

n-1 ≥ λ

n

1

A

L λ1 ≤ λ

2 ≤ λ

3 ≤ … ≤ λ

n-2 ≤ λ

n-1 ≤ λ

n

0

# of 0s = # of components

Page 200: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

Page 201: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

v1

Page 202: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

v1

a

a

a

a

a

a

a

a

a

Page 203: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

v1

a

a

a

a

a

a

a

a

av

2

Page 204: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

v1

a

a

a

a

a

a

a

a

av

2

.37

.37

.28

.28

.12

-.23

-.38

-.38

-.43

Page 205: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

v1

a

a

a

a

a

a

a

a

av

2

.37

.37

.28

.28

.12

-.23

-.38

-.38

-.43

Page 206: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

v1

a

a

a

a

a

a

a

a

av

2

.37

.37

.28

.28

.12

-.23

-.38

-.38

-.43

Page 207: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

v1

a

a

a

a

a

a

a

a

av

2

.37

.37

.28

.28

.12

-.23

-.38

-.38

-.43

Page 208: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

v1

a

a

a

a

a

a

a

a

av

2

.37

.37

.28

.28

.12

-.23

-.38

-.38

-.43

Page 209: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

Page 210: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

+

Page 211: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

+-

Page 212: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

Page 213: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

2nd Eigenvector

Page 214: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

2nd Eigenvector

3rd Eigenvector

Page 215: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

Page 216: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

Page 217: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Eigenvectors

Page 218: Network Analysis Graphs as Matrices - Michele Coscia · Graphs as Network Analysis Matrices Michele Coscia & Luca Rossi ITU København. Background: Background: Looking at a problem

Textbook Chapters● Adjacency Matrix

– Section 2.4● Stochastic, Graph Spectrum &

Laplacian– http://www.michelecoscia.com/?

page_id=1488● Done with Chapter 2!

– Section 2.11