Professor Hollander begins his presentation with the importance of networks in todays life.

Networks has it uses in our daily life such as electricity,phone and transport network. They are generally
highly complex as they are huge in size, intricate in structure, very dynamic and often overloaded as we
like to use thigns to its maximum capacity. Which makes it sometimes unpredictable as overusing will
make it more volatile or even vulnerable.
There is a major problem in complex network that major companies have set up a resilence action
initiative to deal with complex network. He believes that mathematical can help to understand, model,
control and optimize a network. Mathematics can make a problem more abstract and allow one to think
about it in more abstract terms to form new radical ideas for complex network.
He believes that a combination of stochastics and algorithmic can be used to solve the complex network
problem. Stochastic means to model a complex network as something random. He believes randomness
is a way to model complexity. And algorithmic to allow faster optimization and improve the model process.
The kinds of network that are known are networks such as information networks which includes the
internet, technological networks such as banks. Television and radios, social network such as facebook,
or twitter where theres human interaction, biological networks such as genetics, neural and human brain.
A characteristic of these kind of network are its small scale and scale free.
Small scale means that each node in the network are usually connected to each other in such a way that
the degree of separation between them is usually less than 6. The distance between each node may be
small compared to the size of the network. scale free refers to that at which size you look at the network,
they are operating in the same way.
He gave examples such as truck network where places are nodes and the connecting roads are the
distance between them, traffic such as airplanes who have their defined air roads, mobile telephones
which are mapped by landline. These large networks usually suffer from congestion, despite a fluctuating
demand and they can only be controlled locally. Network can only be optimized locally but not globally.
The necessary information to make network robust and optimal is not available yet.
Mathematics can attempt to model this through the use of random graph. Nodes are plotted and lines are
used to connect the nodes. The nodes can be used to represent many things such as airport, power
generator. Lines are drawn connecting the nodes whenever there is something going on between them,
e.g roads between airports, streets where the cars needs to travel or power line for transmission of
electricity.
Graph is a collection of points, interconnected by lines which can be represented as G(V,E).
Where V is the collection of points and E is the collection of lines. To model a complex network, G is
chosen in a random fashion subject to the constraint, F(k) = k- . F(k) represents the fraction of points with
k lines. The k represents the number of neighbor a point have and is a characteristic exponent which
expresses how far points can be connected to other points. A real network usually have a characteristic
exponent of between 2 to 3. This properties causes the small world property where everybody is highly
connected. The presence of hub is a consequence of dynamic growth of network. Network grows by a
rule by preferential attachment. For example a new website is usally connected to a highly visible network
available compared to a vertex that does not have any connection. This is used to explain universal
properties in the architecture of network
The random model can be complicated with different type of lines, for example, different color, green line
makes it easier to travel than red line which can be used to express complexity for sending messages
along a line. He gave three example, Transport

3 billion people in the world use internet, Singapore 4million user. Singapore not sticking out so much
about 160,000 website.
Facebook have more than a billion active user 60% population in Singapore are actively using.
Network uses in electricity, phone, network, transport.
Networks are generally highly complex
-

Huge in size, intricate in structure( we dun even know what the structure looks like sometimes
Very dynamic, often overloaded. We like to use it maximum capacity means we always push it to
its boundary
Sometimes unpredictable. Overusing makes it prone to be volatile or even vulnerable

Major problems need better way to deal with complex network.

-

Build networks with more resilence, less rigid, etc

How can mathematics help?

1. Understand, model, control, optimize
2. To form radically new ideas for complex networks.
3. Abstract certain problem and think about it in more abstract terms.
Combination of
1. Stochastics = art of hazard
a. A way to model a complex network as something random
b. Randomness is a way to deal with complexity
2. Algorithmics = art of computations
a. To optimize model etc need very fast algorithm to improve modelling process
Most promise to make new contributions for complex network.
What sort of netowkrs
1. Information networks
a. Internet/www
2. Technological networks
a. Banks/television/radios
3. Social networks
a. Facebook/twitter
4. Biological networks
a. Genetics/neural/human brain
Often these networks are
small world they are very highly connected. To go from one part to the other may not have to
travel fast. Distance between network are small compared to size of network. 6 degree of separation.
Most network are usually within 6 connection away from each other.
scale free means if you look closely or further away, they look essentially the same. Network
have fractal properties. They have many scale. E.g. city, countries, continent,
Examples
Known networks

Transport: truck network (roads)

Traffic: airplanes (air roads)
Communication: mobile telephones (landline)
Energy: electricity pylons need to know how much is needed at what time, cannot store.
Making such network is rather challenging
Large network suffer from congestion, despite a fluctuating demand and can only be controlled locally.
Can only be optimized locally but not globally.
Many networks we do not really know what they are looking like.
The necessary knowledge to make networks robust and optimal is not yet available
Theory lags behind practice.
Mathematics can do modelling by doing random graph
The power of abstraction: random graphs
-

Many nodes are connected by many edges

nodes could mean different things in mathematics e.g airport,electricity.power
edges are drawn whenever there are something going between the nodes e.g routes between
airplane, street where cars travel or power lines for electricity
on nodes and lines, things can happen

Theory of graphs
1. graph is a collections of points, interconnected by lines:
2. G=(V,E),
3. V=collection of points, E=collection of lines
Models of complex network, G is chosen in a random fashion subject to the constraint
F(k) = fraction of points with k lines ~= k-tau most network are characterastise by this fraction
where tau is a characteristic exponent expresses how far points can be connected to other points.
( certain percentage of points have 2 neighbours and percentage of points have 3 neighbours), the
random outcome of G is denoted by Gtau.
for many real network tau is usually (2,3)
-

e.g plot of f(k) for the collaboration graph in mathematics tau~2.8

this properties causes the small world property where everybody is highly connected.
Presence of hub

Presence of hubs is a consequence of dynamic growth of networks:

Network grows by a rule by preferential attachment! E.g few hundred users to few billions now.
e.g a new website is usually connected to a highly visible network available compared to a vertex
that does not have any connection.
This is used to explain universal properties in the architecture of networks.
e.g Random graph with different types of line e.g different color, green line easier to travel then red line
can be used to express complexity for sending messages for example.

Three examples
1. Transport
a. With each line e (- E we associate a random time W. which can represent the time that
is needed to move along the line e
b. A quantity of interest is the typical minimal travel time in G which is given by the formula
c. T(G)=38:14
d. Pick 2 points completely at random then pick look at all possible path from v1 to v2 and
sum of the travel time that lies on the path then divide by all the possible choice
2. W(e):
a. Transmission time of a message
b. Cost of a transport
c. Proximity of a friendship
3. T(G):
a. Distance travelled within a given time
b. Range within a given budget
c. Information received from friends
4. The behavior of T(G) depends on the architecture of network, in particular the parameter tau if
G=Gtau
5. T(G) ~ mod(V)1/d, if G=Zd

Show why complex network works

2nd example
3. Virus
a.
b.
c.
d.
e.
f.
g.
h.
i.

Each point vV can be infected or health

Point change their state in a random fasion
Infected -> health at rate 1
Healthly -> infected at rate lamda x number of infected neighbours
Lamda (0,inf) as a parameter
For every G there exist a lamdac(G) [0,inf] such tat
Lamda<lamdac: no epidemic
Lamda>lamdac: epidemic
Can quantify this threshold

4. Vulnerability
a. Suppose that Gtau percolates, ie a positive fraction of points is interconnected, a terrorist
kills all the points with degree >=k*
Often we dont know the architecture
5. does the architecture show itself through behivour of the network through extreme conditions?
a. We can sometimes find out how the network fix together.
b. No theory for this yet
6. Is transport a long a dynamic network more efficient along a static network
a. Can network be made more effective by making network more dynamic
7. Can the breakdown of a network of a network be predicted by doing online measurements and
thereby be avoided?
8.
Ultimate goal
Build self orgainising network
Develop a universal theory for intelligent network
Future challenges
Social networks
Neural networks

Everywhere in the world people are connected via networks. Think of Internet, Facebook and Twitter, but
also of road traffic, transport of merchandise, mobile telephones and electricity grids. Such networks have
become indispensable to our modern society. However, they are generally very complex: huge in size,
intricate in structure, very dynamic, often overloaded, sometimes unpredictable, and at times even
vulnerable. This is worrying. To better understand complex networks, to model them adequately, and to
control and optimize them in an efficient manner, new ideas are required. Mathematics is a powerful tool
that has lots to offer. In this talk I describe a few examples of complex networks, discuss a few key
questions, and give an impression of what mathematics is able to do. The combination "stochastics" (the
art of hazard) and "algorithmics" (the art of computation) form the basis of a new perspective on networks.
The ultimate goal is to design and build intelligent networks. The talk is aimed at a non-mathematical
audience with an interest in science.