Complex systems 3

Preview:

Citation preview

Complex Systems

Session 3.

Self-organization and criticality

The Matthew Effect

Robert K. Merton (1968)

Herbert A. Simon (1955) Nobel M. Prize (1978)

Udny Yule (1925)

Yule-Simon process and distribution

SzEEDSM Complex Systems 2017

Yule-Simon

SzEEDSM Complex Systems 2017

Preferential attachment

SzEEDSM Complex Systems 2017

Scaling in networks

SzEEDSM Complex Systems 2017

Complex Networks

SzEEDSM Complex Systems 2017

Degree distribution of Internet

routers

SzEEDSM Complex Systems 2017

α= 1.1

Degree distribution of WWW pages

α= 1.1

Degree distribution in Social

Networks

Gibrat’s law

Robert Gibrat (1931)

Growth rate (x=company- asset- or city size)

Growth is independent of size

Under this assumption, size distribution is power law again

with α approximately 1.

SzEEDSM Complex Systems 2017

Firm size distribution

SzEEDSM Complex Systems 2017

Sales &

Profit

SzEEDSM Complex Systems 2017

The Black Swan

SzEEDSM Complex Systems 2017 Nassim Taleb

Fat Tail

SzEEDSM Complex Systems 2017

Earthquake size distribution

SzEEDSM Complex Systems 2017

Phase transitions

Ferromagnetism and the Ising

model

Cluster size distribution

Mean field approximation

Mean field diagram

Bistability and bifurcation

Percolation

SzEEDSM Complex Systems 2017

Epidemic processes

SIS model

Traffic congestion

Inverse U curve

Social collapse

Collective intelligence

Power outages

10 4 10 5 10 6 10 710

-2

10-1

100

101

N= # of customers affected by outage

Frequency

(per year) of

outages > N

1984-1997

August 10, 1996

Square

site

percolation

or

simplified

“forest

fire”model.

The simplest possible toy model of cascading

failure.

connected

not

connected

Connected clusters

A “spark” that hits

a cluster causes

loss of that

cluster.

yield

=

density

- loss

Assume: one randomly

located spark

(average)

yield

=

density

- loss

Think of (toy) forest fires.

(average)

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

(avg.)

yield

density

“critical point”

N=100

Critical point

criticality

This picture is very generic.

100

101

102

103

104

10-1

100

101

102

Power laws Criticality

100

101

102

103

104

10-1

100

101

102

Power

laws: only

at the

critical point

low density

high density

Life, networks, the brain, the universe and

everything are at “criticality” or the “edge of

chaos.”

Does anyone really believe

this?

Self-organized criticality:

dynamics have critical point as global attractor

Simpler explanation: systems that

reward yield will naturally evolve to

critical point.

Would you

design a

system this

way?

Maybe random

networks

aren’t so great

High yields

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

isolated

critical

tolerant

Why power laws?

Almost any

distribution

of sparks

Optimize

Yield

Power law

distribution

of events

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

random

“optimized”

density

yield

5 10 15 20 25 30

5

10

15

20

25

30

Probability distribution (tail of normal)

High probability region

Optimal “evolved”

“Evolved” = add one site at

a time to maximize

incremental (local) yield

Very local and limited optimization, yet still

gives very high yields.

Small events likely

large events

are unlikely

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

random

“optimized”

density

High yields.

Optimized grid

Small events likely

large events

are unlikely

Optimized grid

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

random

grid

High yields.

This source of power

laws is quite universal.

Almost any

distribution

of sparks

Optimize

Yield

Power law

distribution

of events

Tolerance is very different

from criticality.

• Mechanism generating power laws.

• Higher densities.

• Higher yields, more robust to sparks.

• Nongeneric, won’t arise due to random

fluctuations.

• Not fractal, not self-similar.

• Extremely sensitive to small perturbations that were

not designed for, “changes in the rules.”

Extreme robustness and extreme hypersensitivity.

Small

flaws

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Recommended