AITIC Session 05

The architecture of complexity (BARABÁSI 2007)

Neo-

Second Law of Thermodynamics

Computational Austrian
First law of Thermodynamics

classical

Heisenberg’s Uncertainty Pr.

Economics
Newton’s Laws of motion

Economics Simulation
Mathematical (agent models)
Game Theory
Modelling
DEDUCTION
Complexity
Statistical Science
Inferences INDUCTION

Artificial Heuristics
ABDUCTION Intelligence
1
Barabási A-L (2007) The architecture of
complexity. IEEE Control Systems Magazine,
(August), 33–42

2
Architecture of Complexity: Barabási (2007)
• Complex Systems
• are not made of identical components
• interactions among the components are manifestly
nonlinear
• unpredictable behaviour can emerge from the nonlinear
interactions of a few components
• Yet they display signatures of order and self-organization
• In complex systems created by humans, interacting nodes form networks, where
(usually) each node interacts with only a small number of selected partners
whose presence and effects may be felt by far away nodes. Examples: WWW,
Internet, Power Grids, Transportation systems
• A complete theory of complexity does not yet exist
3
Network Topology: Random (or exponential)
• In random (or exponential) networks, N nodes are connected by L
randomly placed links.
• Probability that a node is connected to k other nodes decreases
exponentially for large k
• Most nodes have the same number of links
• Nodes with large number of links are almost always absent
• Absence of nodes with significantly more links than k imposes a
natural scale in the network
• Ex.: Road network connecting cities in the USA
• Add a city only 2-5 links increase
4
Network Topology: Scale Free Networks
Complex networks created by humans often have an
underlying architecture guided by universal principles
• For ex., the World Wide Web (WWW) (and the Internet) is
dominated by a small number of nodes that are highly connected
to other nodes (others have much fewer connections) => Scale
free n/w, follows Power law distribution
• Commercial Airline transportation in US EU has a few
densely-connected hubs (Chicago, Frankfurt)
• Add a non-hub city 1 to 3 links increase
• Add a hub : 100 links shall increase
5
 Difference: WWW and the Internet
• The WWW contains several billion documents (Web pages), which
represent the nodes of a complex network.
• These documents are connected by uniform resource locators (URLs), which
are used to navigate from one document to another
• Since the WWW is a directed network, each document is characterized by the
number k out of its outgoing links and the number k in of its incoming links
• Unlike the WWW, the Internet is a physical network whose nodes are
routers and domains and whose links are the phone lines and optical
cables that connect the nodes
• WWW continuously expands by adding new Web pages, while the
Internet grows with the installation of new routers, computers, and
communication links
6
 Small World structure
• Draws from the fact that: any two persons are typically five to six
handshakes away from each other
• For the 800 million nodes of the WWW in 1999, the typical shortest path
between two randomly selected pages was around 19
• for 200 million nodes this distance is 16
• The WWW displays a high degree of clustering
• the probability that two neighbours of a given node are also linked is much greater
than the value expected for a random network

A small-world network is a type of mathematical graph in which most nodes are not neighbours
of one another, but the neighbours of any given node are likely to be neighbours of each other
and most nodes can be reached from every other node by a small number of hops or steps.
(Source: Wikipedia) 7
 Preferential Attachment

• A person is more likely to link a Web page to highly connected documents

on the WWW, whose existence is well known
• Network engineers tend to connect their institution’s computers to the
Internet through high-bandwidth nodes, which inevitably attract a large
number of other consumers and links.
• Growth and preferential attachment jointly lead to the appearance of the
hub hierarchy that exemplifies the scale-free structure
• rich-get-richer positive-feedback phenomenon
A preferential attachment process is any of a class of processes in which some quantity, typically some form of
wealth or credit, is distributed among a number of individuals or objects according to how much they already
have, so that those who are already wealthy receive more than those who are not . (Source: Wikipedia)
8
Robustness & Vulnerability of scale-free n/ws
• Scale-free networks cannot be broken into pieces by the random
removal of nodes
• Since there are far more weakly connected nodes than hubs, random removal
most likely affects the less-connected nodes. The removal of a node with a
small degree does not significantly disrupt the network topology
• Scale-free networks are especially vulnerable to targeted attacks
• Removal of a small fraction of the most connected nodes (hubs) might break
the network into pieces. Well-informed hackers can easily handicap the
network by targeting hubs for attack
• The critical threshold of a scale-free network is zero. Therefore, all viruses,
even those that are only weakly contagious, eventually spread and persist in
the system.

9
Human Dynamics
• Timing of many human activities, ranging from communication to
entertainment and work patterns, follow non-Poisson statistics
• There are bursts of rapidly occurring events separated by long periods
of inactivity (for example this is seen w.r.t. email communication)
• When individuals execute tasks based on a perceived priority, the
tasks’ timing follows a heavy-tailed (Pareto) distribution, with most
tasks being rapidly executed while a few tasks experience long waiting
times
• High-priority tasks are executed soon after they are added to the list, while
low-priority items must wait until all higher priority tasks are cleared. This
protocol forces lower priority items to stay on the list longer than higher
priority ones.
10