Chapter 1

Setting the Stage: Complex Systems,

Emergence and Evolution

1.1 Complex and Adaptive Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 What Is Chaos? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Constructing Artificial Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Importance of Emergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Dynamic Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6 Is There Evolution at Work? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6.1 Adaptation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.7 Distributing Intelligence? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.8 Modeling and Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.8.1 Research Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Natural Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Control Engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Cellular Automata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Agent-Based Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

COMPLEX SYSTEMS are composed of many interconnected elements, work-

ing individually, but producing an overall global system behavior. The funda-
mental desire to study how these complex systems behave comes from various
multifaceted disciplines such as biology, economics or even social sciences.
Examples include large ant colonies (composed of individual ants cooperat-
ing to exploit available food sources), the human nervous system (composed
of tiny neurons sending and receiving signals in the human body) or social
structures (such as communication networks). Depending on the system be-
ing studied, individuals behave in organized (or disorganized) ways, leading
to unpredictable overall system behavior. This phenomenon, referred to as
emergent behavior, is a direct consequence of individual behaviors inside the
system and their interactions among each other.
Engineering projects have taken inspiration from complex natural systems
to build better and reliable infrastructures. Understanding how cities survive
and how crowds behave are key elements in designing buildings or studying
how economies work.
Agent-based modeling (ABM) is a unique modeling technique that allows
a one-to-one mapping to natural systems. Modelers understand complex sys-
tems, how they are composed of multiple individuals and their interactions

1
2 X-Machines for Agent-Based Modeling: FLAME Perspectives

"The whole"
emergent, global structure

"Parts"
Local interaction

FIGURE 1.1: Emergence in complex systems. cf. [116].

among themselves and the environment. Writing agent-based models draws

inspiration from parallel computation, software engineering, data analysis and
simulation, to achieve reliable simulation models as virtual complex systems.
This book aims to study and provide readers with principals involved in build-
ing and writing agent-based models from a software engineering perspective.
Presenting itself primarily as a modeling and simulations tool, the book cov-
ers computational challenges of software engineering, parallelization, verifica-
tion and validation, all of which are issues for computer and other scientists
when developing reliable agent-based models. To explain details from a soft-
ware engineering perspective, we focus on an established agent-based modeling
framework, FLAME, as a guide to understand and build ABM approaches. By
discussing the range of projects and computational complexities it has faced in
the research area, various computational challenges are discussed from model
conception, building, execution and testing, in fields of biology, social networks
and economics.
Complex systems are studied in two ways - either as one collective sys-
tem, or as a collection of individuals interacting with each other to produce
an overall behavior. The dynamic individual behavior can be studied using
mathematical formulas [100] such as differential equations or time-based ac-
tivities. However, using mathematical equations often restricts models to cer-
tain levels of complexity and data being collected. For instance, hierarchical
relationships observed at macro system level, as well as at micro internal level
within individuals cannot be easily studied using equations (Figure 1.1). In
complex systems, local individual interactions cause emergent system quali-
ties at higher levels, allowing emergence to be a consequence of what happens
within these micro levels [114].
 Setting the Stage: Complex Systems, Emergence and Evolution 3
TABLE 1.1: Examples of research carried out in complex
adaptive systems. Adapted from Schuster [171].
Research Area Researchers Year
Darwinian evolution Smith and Szathmary [189] 1995
Chemical networks Kauffman [103] 1993
Ecological networks Sigmund [177] 1993
Insect colonies Bonabeau and Dorigo [25] 1999
Immune system Segel and Cohen 2000
Nervous system Kandel [101] 2000
Economic networks Lane and Durlauf [11] 1997
Social networks Frank [67] 1998
Communication networks Barabasi [4] 2000
Transportation networks Narguney [137] 2000
Evolutionary games Hofbauer and Sigmund [85] 1998

1.1 Complex and Adaptive Systems

Multiple disciplines use complex systems to explain unusual phenomena
and systems characteristics by artificially creating large simulated systems
modeling real systems, aiding understanding on how these systems behave.
Table 1.1 discusses some of the early examples in various disciplines and com-
plex system modeling. Some of the common features of these systems are
summarized in Figure 1.2.
Individual elements exist on multiple levels within the system, allowing
hierarchies, and even hierarchies, to develop where two systems are mutually
exclusive and continuously interacting. These elements can act as representa-
tives of either a single performing individual or as a collection of individuals
such as groups of multiple individuals. Each element evaluates its behavior
based on a reward system and adapts to perform better in the current sys-
tem conditions. The reward system is determined by a performance measure,
where individuals use receptors to read signals and functions to assess these
performances.
Adaptiveness of elements is a unique feature that complex systems pos-
sess. Researchers have studied how systems predict and readjust efficiently to
changing conditions. For example, Hopfield [91] showed that system adaptive-
ness is highly influenced by presence of noise and attractors in the system.
Attractors are environmental points that cause elements to deviate from their
ideal paths of behavior, possibly when systems start to show chaotic behaviors.
4 X-Machines for Agent-Based Modeling: FLAME Perspectives

Key:
Type of complex system The kind of scenarios
(Function of the system) (Agents)

Prebiotic Evolution Chemical Networks

(Emergence of life) (Molecules)

Darwinian Evolution Genetic Algorithms

(Formation of species) (Genes)

Brain Evolution Neural Networks

(Learning, predictions) (Neurons)

Social Evolution Evolutionary Games

(Ecologies, economies) (Players)

Common Attributes:

Memory,
Computation Reward system, Noise and
Storing
input, action Assign credits control
information

FIGURE 1.2: Examples of complex adaptive systems, their models and com-
mon characteristics. cf. [171].

1.2 What Is Chaos?

In mathematics, chaos theory is the description of a dynamic system that
exhibits high sensitivity to initial conditions of the system. Conversely, chaotic
behavior, in common language, also translates into an unpredictable or unper-
ceived behavior. Chaos, thus, has multiple meanings depending in the context
it is used. In this book, a chaotic effect refers to an emergent behavior which
is unpredictable, or otherwise unknown to observer at the beginning of the
simulation. There is a separate research field which involves measuring chaotic
points or attractors in a system during simulations, usually measuring initial
conditions and then comparing them to a series of outputs generated. Testing
these effects of chaos theory is out of the scope of discussions presented here.
Complex systems are known to sometimes go into chaos. Derived from
ancient Greek [145], it describes a state that lacks order or even predictability.
Langton [114] coined the term ‘edge of chaos’, which was used to describe
 Setting the Stage: Complex Systems, Emergence and Evolution 5

the point at which system starts exhibiting chaotic behavior, or the point
at which it becomes extremely sensitive to initial conditions. This sensitivity
sometimes produces bifurcations (or branches into two possible behaviors)
that are difficult to predict (Figure 1.3).

FIGURE 1.3: Bifurcation diagram in a logistic map. Adapted from [122].

1.3 Constructing Artificial Systems

Complex systems can be seen as large problems that can be solved as
collections of smaller problems. For instance, ant colonies and individual ant
behavior are being studied to give possible solutions to computer networking
problems [170], or understanding how prices behave in stock markets.
Large engineering applications, also made up of tiny parts working to-
gether, can have precise predictable behavior. These individual units always
perform as they ought to, unless they fail due to some dependencies which were
difficult to predict. Economic systems also exhibit a wide variety of emergent
behaviors, with humans sometimes not paying their credit bills regularly or
buying houses without paying mortgages, cited as some of the reasons for
2008 credit crunch [33]. This unpredictability and randomness of individu-
als, leads to the failure of large systems performing as predicted. The extent
of failure, having a domino effect on surrounding elements, depends on how
many individuals deviated, allowing complex systems research to become a
multi-dimensional problem with techniques from psychology and behavioral
economics, enhanced by methods in computer science.
6 X-Machines for Agent-Based Modeling: FLAME Perspectives

1.4 Importance of Emergence

Goldstein [74] argues that “emergence refers to rising of novel and coher-
ent structured patterns and properties during the process of self-organization
of complex systems”. But Anderson [8] points out that due to scale and com-
plexity, it is not necessary that the built model would always turn out to be
same as its individual real parts. This notion leads to the fact that emergence
itself cannot be defined as a perfect pattern with multiple result possibilities.

“The ability to reduce everything to simple fundamental laws does

not imply the ability to start from those laws and reconstruct the uni-
verse. The constructionist hypothesis breaks down when confronted with
the twin difficulties of scale and complexity. At each level of complexity
entirely new properties appear. Psychology is not applied biology, nor
is biology applied chemistry. We can see that the whole becomes not
merely more, but very different from sum of its parts.” [8]

Modeling a system is the process of creating a replica of the system. This

could be done by considering only a few aspects of what is needed to be ob-
served from that system, or what modelers desire to test. For instance, testing
small gears working together in a clock could either be tested with individual
elements modeled as gears, or whole collection of gears connected to the nee-
dle, taken as one individual. Modeling depends on modeler requirements to
how they want to represent the system.
The model would also be simulated a number of times to understand its
average behavior. Randomness in complex systems can sometimes lead to
unpredictable patterns, which makes testing a concrete part of modeling.

1.5 Dynamic Systems

Complex systems can adapt to changing environmental conditions. Their
ability to cope with changes and their survival makes systems extremely robust
and favorable for inspirations in engineering applications. Traditionally, nu-
merical equations with differentiation are used to represent dynamic systems
as functions with respect to time. Examples of such equations are Newton’s
law of motion for particles and forces, represented as expressions of veloc-
ity, acceleration during movement and direction of travel for particles. The
Navier-Stokes equations are used to describe motion of fluid substances, used
 Setting the Stage: Complex Systems, Emergence and Evolution 7

to model behavior of water in pipes. Using these equations can represent the
system as a derivative of time [144],
dx
Ẋ = = F (x) (1.1)
dt
Equation 1.1 shows change in a system represented with time, where X =
(x(1) , x(2) , ..., x(k) ) and k is the number of states of the system. The system
state is given as a property (for all elements) as a snapshot at the time. This
can include individual element properties, environmental conditions and any
other attributes involved. Basically, it is a snapshot of the system at time t,
taken between starting time t = 0 and time t = k. In this way, it is possible to
determine how the system looks at time t = t + 1, if state at t = t is known.
However, complex systems are emergent systems. This makes it sometimes
difficult to predict or anticipate, how the system would look at t + 1 as there
are too many individual element interactions leading to its snapshot at t + 1
due to randomness in individual behavior. These systems are also irreversible,
which means it is also difficult to work backwards and deduce what the past
state was even if current and future states are known. Researchers can deduce
a number of reasons why the system behaved in the way, by running repeated
simulations and testing the effect of all elements on overall system behavior.

1.6 Is There Evolution at Work?

Being continually adaptive, systems show continuous dynamic change.
This uses fewer or basic starting conditions and assumptions to grow into
complex system behavior. As time moves forward, certain conditions can be
changed to alter its behavior and future system states. Other components that
play a key role include geography or locations in the system. Geography can
influence in ways such as the following:

Communication span for each individual. Messages or communication

between individuals, which are limited to particular individuals in an
area. This gives them more information and act accordingly.
Messages influence personal behavior. Received messages can be used
to determine the next strategies to play based on the incoming informa-
tion.
Influence of resource availability. Depending on their locations, each in-
dividual has various levels of resource available, that can affect its behav-
ior. For example, in a ant colony model, if an ant comes across a stream
of water, it can locally change its on-course path, effectively adapting to
the situation and locality. Over time, the system will display a stronger
8 X-Machines for Agent-Based Modeling: FLAME Perspectives

ant path being created, that deviates from water. This is an important
ability to ensure survival in changing environments.

1.6.1 Adaptation
Adaptation is the ability of individuals changing their behavior or functions
to survive better in their present conditions. Examples include developing
ability to run faster or hiding from predators. In evolution, organisms with
successful adapting capabilities will grow, improving on their likelihood to
survive.
Karl Sims [181] presented his work on artificial life, where he displayed
evolution in action by creating a computer simulated block creature that had
rectangular blocks hinged together. Each block was flexible and allowed to
move, such that the creature could restructure itself to suit to the environment
it was in.

FIGURE 1.4: Examples of Karl Sims’s creatures. cf. [180].

The creatures were evolving towards a common goal, which was to swim
as fast as they could through a water environment (Figure 1.4). Simulated
results showed that there were no optimum solutions, as creatures could not
understand water mechanics and its behavior. However, the results showed
new designs quickly generated, increasing the rate of survival for creatures in
water. Sim’s experiments showed evolution was at play when performed with
particular goals for survival.
Through various successful adaptations, an emergent behavior can be ob-
served, seen as an outsider view of the system. This is known as evolutionary
drive in the system, as conditions and time force individuals to change.
Evolution is a term borrowed from biology, where organism populations
adapt from one generation to the other. Over time, generations accumulate
 Setting the Stage: Complex Systems, Emergence and Evolution 9

various differences with each other, depending on how they adapt, allowing
new generations to have a gradual divergence from the starting pool of char-
acteristics. These differences are brought about due to locations, limited in-
formation or available resources, allowing emergence of new species that can
better adapt to given situations. Some individuals not well adapted, will grad-
ually die out, leaving only strong ones to multiply.
The term reliability refers to probability of a component operating sat-
isfactorily during a certain time frame. Quantifying reliability requires one
to define, assess and combine probabilities of risk and system behaviors [27].
This may require identifying system variabilities and vulnerabilities, to predict
lifetimes to assess model reliability.

1.7 Distributing Intelligence?

Evolution is learning and not intelligence. Minsky [134] claimed that in-
telligence is used to emphasize swiftness and efficiency of a solution.

“Evolution’s time rate is so slow that we don’t see it as intelligent,

even though it finally produces wonderful things we ourselves cannot yet
make”.

Evolutionary behavior can be observed at multiple levels. Every layer can

be ‘zoomed in’ to see different patterns of behavior emerging. Johnson [98]
discusses an example of a city as a complex system, where the city itself
behaves like one individual system, consisting of a number of thriving neigh-
borhoods within. Each neighborhood consists of a collection of people involved
in complex networks such as traffic networks. Similar to ant colonies, a city
is a system which has decentralized control, learning from local interactions
making a man-made self-organising system using emergence.
However, the beauty of these systems lies in the individual elements. These
units can think, restructure and communicate with other units. Some of these
characteristics are summarized as follows:
System is part of a larger complex system. The systems are connected
to other systems as a hierarchy or using input or output branches.
Systems are open systems. The systems are interacting with other sys-
tems continuously with no bounds. In a closed system, the system exists
as an isolated entity with specific boundaries, like gas molecules con-
tained in a container, where conditions of thermodynamics hold. En-
tropy changes can therefore be predicted. However, some systems are
10 X-Machines for Agent-Based Modeling: FLAME Perspectives

a collection of various smaller systems intertwined, have to modeled as

open systems.
Systems are dynamically changing. The environment around the indi-
viduals is constantly changing, influencing their behavior.
Display emergent behavior. The emerging behavior can be studied at
macro levels. For instance, insect colonies achieve their goals quicker
by working collusively among individual ants.
Individuals are adaptive. Depending on the changing environment and
available system resources, individuals adapt their behavior to survive
in given conditions.
Individuals are selfish. All individuals work for their own benefit using
local information.

1.8 Modeling and Simulation

Modeling and Simulation (M&S) is a core research area under scientific
computing, where artificial systems are created as models and simulated in
a virtual environment. Executing them in a virtual environment allows to
safely assume changes, in order, to predict how the system would behave
when certain changes are introduced in real world situations.
However, it is important to note that a model is only an approximate rep-
resentation of the system, showing only basic functionalities being explored.
It is often a very simple representation of the system, with clearly defined
assumptions embed into the model while it is constructed.
A model is a representation of an object, a system or an idea represented
in a form other than that of the entity itself [175]. Simulation allows the model
to be tested in a virtual world to check its reaction to certain conditions. The
model’s design would ensure how reliable it is for making predictions.
There are multiple forms in which models are created, such as physical,
where models are constructed as prototypes, or scale models, where they repre-
sent systems, and mathematical, where models are constructed as analytical
mathematical notations, linear and simulation-based representations. In all
cases, techniques chosen to construct models, depend on the objectives and
aims of the modelers. Model examples include, but are not limited to,
• Engineering applications: Test if certain temperatures will affect smooth
running of the engine. These include examples from designing and ana-
lyzing manufacturing systems or transport systems.
 Setting the Stage: Complex Systems, Emergence and Evolution 11

• Biological models: Of tissues, neurons or cellular models to study effects

of chemicals and drugs on cell behavior.
• Economic models: Of various markets such as stock markets, labor mar-
kets or economic systems to study the introduction of migration, taxes
and money on the overall market behavior.
• Social science models: To study effects of various population dynamics
on areas and resources.
• Evaluating systems: Hardware or software performance for a computer
system or new military weapons system or tactics on enemy forces.
• Designing communications systems and message protocols for commu-
nicating entities.
There are various steps involved in constructing M&S mapping from real
world situations and simulating them in a virtual world. The steps involved
are as follows:
Step 1. Identify problem being investigated in real world: This is
very specific to hypothesis being tested, which can usually not be tested
in real or natural settings due to costs or impacts. This justifies it being
tried out as a virtual experiment first.
Step 2. Formulate model problem: Formulate a model for a system in a
manner by which it is created as a virtual representation. This involves
determining assumptions of the model, hypotheses being tested, kind of
data being collected from real world to test it and, finally determining
which tools to used to create the model. This usually involves talk-
ing to domain experts and collecting relevant data to construct most
accurate system representations. Computer simulations involve multi-
disciplinary approaches, where computer scientists work with biologists
or economists to construct computational models for systems from their
disciplines. A computer scientist has to ensure the model has been cor-
rectly represented and all necessary behaviors are captured by it.
Step 3. Simulate model using relevant software toolkits: Use software
tools to simulate a model.
Step 4. Analyze data collected: The simulation results are collected and
analyzed. The results can be used to find discrepancies and test theory
predictions, allowing modelers to verify their models.

Step 5. Data mining techniques: Data analysis techniques such as ma-

chine learning, pattern findings and data visualizations help determine
the simulation conclusions, in terms of testing the hypothesis.
12 X-Machines for Agent-Based Modeling: FLAME Perspectives

Step 6. Validity and verification of model: Involve validating and ver-

ifying results of simulation, to test if they are correct for conclusions
being drawn on hypotheses. At this step, review of the model correct-
ness and result reliability can circle back to step one, by finding issues
or wrong assumptions in the initial model constructed.
It is important for developers and researchers to remember that a model is
not a goal of the experiment, but it is a process by which simulation will find
solution to the hypothesis being tested. Thus the model is only an enabler to
the process being investigated [59]. Figure 1.5 shows a flow chart of processes
involved when creating biological models. The figure highlights how model-
ers sometimes need to rework through initial model descriptions, to correct
models, after expert advice and results are obtained.

DESIGN: Understand model

Identify the memory variables of the agent.

Identify the functions the agent does.
Identify the communication the agent does with other agents.

Simulated
results
invalid IMPLEMENTATION:
Simulate the model to allow the agents
to interact.

TESTING:
Real experimental
Observe behavioural data using graphical
data
techniques and tools.

Simulated results
validated

Release model
understanding behaviour.

FIGURE 1.5: Modeling process in biology simulations. cf. [107].

1.8.1 Research Examples

Natural Systems
Falling under area of swarm intelligence, ant colonies are extremely efficient
in finding shortest possible routes to food in minimum time. Proposed in
Dorigo’s PhD work [53], ant colony optimization algorithms can solve complex
 Setting the Stage: Complex Systems, Emergence and Evolution 13

problems like the travelling salesmen problem and network routing problems
for dynamic scenarios (Figure 1.6).

FIGURE 1.6: Various time steps showing ant colonies finding and forming
routes to food sources. cf. [23].

Biological systems like the human body are extensively studied as complex
systems. The study of NFκB molecule is an example of studying how tran-
scription factors work in cells [149]. Apart from cells, foreign organisms like
bacteria, living in human bodies, have also been subject of much research,
where bacterial behavior is often studied in human stomachs to determine
how they survive in less oxygen levels [125].

Control Engineering
Control systems engineering involves design of robust applications func-
tioning in real world conditions. Research in this area has grown to accom-
modate various aspects like [95]
• Regulating control of systems.
• Building large systems like bridges or computers.
• Dynamic environmental conditions.
• Optimization and distributing data over large systems.
Being treated as complex systems, systems control theory analyzes large
systems as collections of smaller units working together to produce the system
output. For instance, Figure 1.7 shows a system made of three interacting units
A, B and C. These units can be a capacitor, transistor or a memory chip,
working together in the system. The system output produces an effect that
brings change in system input at next time step. This becomes an important
feedback loop, allowing the system to adapt to changing conditions in the
environment (dynamic environments).
14 X-Machines for Agent-Based Modeling: FLAME Perspectives

Inputs from A Output

the environment
B

Feedback from output

affects inputs

FIGURE 1.7: A system working in the environment. The system is composed

of three elements working together to make the system work efficiently. Output
produces a feedback, that produces change in the system as time progresses.

Cellular Automata
Cellular automata models have stemmed from basics of computational
theory, mathematics and biology. Developed by Ulam and von Neumann [198],
they were able to prove the notion of one robot producing another robot or
also known as ‘the principle of self-replicating systems’. A famous example is
the ‘Game of Life’ by John Conway which uses four simple rules of generations.
Here, every element is treated as a cell that transitions based on strict rules
predefined by life generations [159].
Being used as a more powerful computational model [203], principles of
cellular automata allow individual cells to react and change their states based
on their surrounding neighboring cells. If visualized as a plane of cells, there
can be a pattern that is observed moving across from one point to other, by
subsequent reaction of cells. For example, vibration of molecules in a solid,
when provided with heat, acts as a wave propagating from one point to other.

Agent-Based Models
The word agent has multiple definitions by different modelers. With re-
spect to agent-based models, the following definition is used in this book:

“An agent is a computer system that is situated in some environment,

and that is capable of autonomous action in this environment in order
to meet its design goals” Wooldridge and Jennings [205]

It does not specify that every gear in a clock be modeled as one agent or
the whole unit to be treated as one agent to reach model goals. This allows
modelers to define their own agents and their behaviors per model.
 Setting the Stage: Complex Systems, Emergence and Evolution 15

Ideologies surrounding cellular automata models gave birth to concepts

of agent-based modeling. Reynolds, in 1985, introduced agent-based models
as a driving force for scientific computing, particularly using powerful par-
allel computers. The computer graphics expert produced the boids example
which depicted flocking of birds. Later, Langton coined the term artificial life
to describe similar simulations [201]. These allow simulations of large agent
populations to be executed in controlled environments, examining affects of
various rules on agent interactions.
Agent-based models encourage bottom-up approaches, allowing research to
focus on individual elements interacting with each other, rather than looking
at complete scenarios. Initially, pattern in models was proved using differential
equations with common examples being found in economic modeling, where
mathematical formulas are still being used to prove behavior of ideas. Miller
and Page [131] and Epstein [56] have favored agent-based approaches by saying
that research should be intensified to focus into agents rather than whole
systems, realistically allowing humans to be modeled as agents rather than
differential equations.

FIGURE 1.8: Research areas of ‘Scientific Computing’ and ‘Distributed

Computing’ have a close relationship in agent-based modeling.

Advances in parallel and distributed computing can help scientific compu-

tation as data and computation grows (Figure 1.8). These can allow data to
be processed quickly and analyzed in real time to test models and make better
predictions of real complex systems. This work is considerably helped by com-
puting experts in parallel architectures to work with multi-domain scientists
to hasten scientific discovery in their fields.