systems

Communication
Visual Analysis of Nonlinear Dynamical Systems:
Chaos, Fractals, Self-Similarity and the Limits
of Prediction
Geoff Boeing
Department of City and Regional Planning, University of California, Berkeley, CA 94720, USA;
[email protected]; Tel.: +1-510-642-6000

Academic Editor: Ockie Bosch

Received: 7 September 2016; Accepted: 7 November 2016; Published: 13 November 2016

Abstract: Nearly all nontrivial real-world systems are nonlinear dynamical systems. Chaos describes
certain nonlinear dynamical systems that have a very sensitive dependence on initial conditions.
Chaotic systems are always deterministic and may be very simple, yet they produce completely
unpredictable and divergent behavior. Systems of nonlinear equations are difficult to solve
analytically, and scientists have relied heavily on visual and qualitative approaches to discover and
analyze the dynamics of nonlinearity. Indeed, few fields have drawn as heavily from visualization
methods for their seminal innovations: from strange attractors, to bifurcation diagrams, to cobweb
plots, to phase diagrams and embedding. Although the social sciences are increasingly studying these
types of systems, seminal concepts remain murky or loosely adopted. This article has three aims.
First, it argues for several visualization methods to critically analyze and understand the behavior of
nonlinear dynamical systems. Second, it uses these visualizations to introduce the foundations of
nonlinear dynamics, chaos, fractals, self-similarity and the limits of prediction. Finally, it presents
Pynamical, an open-source Python package to easily visualize and explore nonlinear dynamical
systems’ behavior.

Keywords: visualization; nonlinear dynamics; chaos; fractal; attractor; bifurcation; dynamical

systems; prediction; python; logistic map

1. Introduction
Chaos theory is a branch of mathematics that deals with nonlinear dynamical systems. A system
is simply a set of interacting components that form a larger whole. Nonlinear means that due to
feedback or multiplicative effects between the components, the whole becomes something greater
than the mere sum of its individual parts. Lastly, dynamical means the system changes over time
based on its current state. Nearly every nontrivial real-world system is a nonlinear dynamical system.
Chaotic systems are a type of nonlinear dynamical system that may contain very few interacting parts
and may follow simple rules, but all have a very sensitive dependence on their initial conditions [1,2].
One might expect that any simple deterministic system would produce easily-predictable behavior.
Yet, despite their deterministic simplicity, over time, these systems can produce wildly unpredictable,
divergent and fractal (i.e., infinitely detailed and self-similar without ever actually repeating) behavior
due to that sensitivity. Forecasting such systems’ futures thus requires an impossible precision of
measurement and computation. Chaos fundamentally indicates that there are limits to knowledge and
prediction because some futures may be unknowable with any precision. Further, interventions into
a system may have unpredictable outcomes even if the intervention is very minor, as tiny effects can
compound (or be damped) nonlinearly over time.

Systems 2016, 4, 37; doi:10.3390/systems4040037 www.mdpi.com/journal/systems

Systems 2016, 4, 37 2 of 18

Real-world chaotic and fractal systems span the spectrum from leaky faucets [3], to ferns [4],
to heart rates [5–7], to cryptography [8,9]. Many scholars have studied the implications of nonlinearity,
chaos and fractals for the social sciences, including sociology [10,11], urban studies [12–16],
economics [17–21], architecture [22,23] and city planning [24–27]. One constant throughout the
interdisciplinary history of nonlinear dynamical systems’ study is that nonlinear systems are
extremely difficult to solve analytically because they cannot be broken down into constituent parts,
solved individually, then recombined as a solution. Scientists have instead relied heavily on visual and
qualitative approaches, a perspective first developed by Henri Poincaré in the late 1800s, to discover
and analyze the fascinating dynamics of nonlinearity [28,29]. Information visualization helps analysts
detect and examine hidden structure in complex datasets [30]. In particular, few fields have drawn
as heavily from visualization as nonlinear dynamics and chaos have for their pivotal discoveries,
from Lorenz’s first visualization of strange attractors [31], to May’s groundbreaking bifurcation
diagrams [32], to phase diagrams for discerning higher-dimensional hidden structures in data [33].
Such nonlinear analysis is particularly useful, yet underutilized for exploring time series [34,35].
These methods in turn have broad applicability to visual systems analysis. This article introduces
nonlinearity and chaos interdisciplinarily through the methods of data visualization, using a logistic
model to dissect the terminology, visualize pertinent features of chaos and fractals and discuss
the wide-ranging implications for knowledge and prediction. It makes three primary contributions.
First, it reviews and disseminates advanced visualization techniques for the qualitative analysis of
nonlinear dynamical systems’ behavior to an interdisciplinary body of systems scholars. Second,
it provides a visual introduction to the salient concepts of nonlinearity and chaos to a scholarly
audience. Although the social sciences are increasingly studying these types of systems, some of the
seminal concepts remain murky or loosely adopted in the theoretical literature [36]. Most formal
treatments of chaos and nonlinear dynamics in the scholarly literature are densely technical and geared
toward an exclusive audience of mathematicians and physicists. For this article, rather, readers require
only a familiarity with algebra. We thus do not cover the rigorous mathematical underpinnings of chaos
and nonlinear dynamics, but the references throughout cite both the original foundational publications
in this field, as well as recent scholarly developments. Interested readers will be well-rewarded in
consulting these works. Third and finally, it presents Pynamical, a free open-source Python software
package for the visual analysis of discrete nonlinear dynamical systems. Comparable tools usually
must be developed from scratch or rely on expensive commercial software, such as MATLAB [37].
Pynamical provides a fast, simple, reusable and extensible new means for exploring system behavior.
The following section provides a background to the logistic map and the concepts of system
dynamics and attractors. Then, we introduce several information visualization techniques to
explore qualitative system behavior, bifurcations, the path to chaos, fractals and strange attractors.
We investigate the difference between chaos and randomness before finally visualizing the famous
butterfly effect and discussing its implications for scientific prediction. All of these models and
visualizations are developed in Python using Pynamical; for readability, we reserve the technical
details of its functionality for Appendix A.

2. Background and Model

Edward Lorenz, the father of chaos theory [38], once described chaos as “when the present
determines the future, but the approximate present does not approximately determine the future” [39].
Lorenz first discovered chaos by accident while developing a simple mathematical model of
atmospheric convection, using three ordinary differential equations [31]. He found that nearly
indistinguishable initial conditions could produce completely divergent outcomes, rendering weather
prediction impossible beyond a time horizon of about a fortnight [40].
How can this possibly happen with a simple deterministic system? We will explore an example
using the logistic map, a model based on the common s-curve logistic function that shows how
a population grows slowly, then rapidly, before tapering off as it reaches its environment’s carrying
Systems 2016, 4, 37 3 of 18

capacity [41,42]. The logistic function uses a differential equation that treats time as continuous.
The logistic map instead uses a difference equation to look at discrete time steps [43,44]. It is called the
logistic map because it maps the population value at any time step to its value at the next time step:
xt+1 = r·xt ·(1 − xt ). This nonlinear equation defines the rules, or dynamics, of our system: x represents
the population at some time t, and r represents the growth rate. Thus, the population level at any
given time is a function of the growth rate parameter and the previous time step’s population level.
If the growth rate is set too low, the population will die out and go extinct. Higher growth rates might
settle toward a stable value or fluctuate across a series of population booms and busts.
Chaos can manifest itself in both continuous (i.e., with dynamics defined by differential equations)
and discrete (i.e., with dynamics defined by an iterated map) nonlinear dynamical systems. The logistic
map is a simple, one-dimensional, discrete equation that produces chaos at certain growth rates. We will
explore this in depth momentarily, but first, we use Pynamical to run the logistic model for 20 time
steps (we will henceforth call these recursive iterations of the equation generations) for growth rate
parameter values of 0.5, 1, 1.5, 2, 2.5, 3 and 3.5. Table 1 presents the results. The columns represent
growth rates, and the rows represent generations. The model always starts with a population level of
0.5 and represents population as a ratio between zero (extinction) and one (the maximum carrying
capacity of our system). If we trace down the column in Table 1 under a growth rate of 1.5, we see that
the population level eventually settles toward a final value of 0.333 after several generations. In the
column for a growth rate of two, we see an unchanging population level of 0.5 across every generation.
This makes sense in the real world: if two parents produce two children, the overall population will
neither grow nor shrink. Thus, a growth rate parameter value of two represents the replacement rate.

Table 1. Population values produced by the logistic map with 7 growth rate parameter values over
20 generations.

Generation r = 0.5 r = 1.0 r = 1.5 r = 2.0 r = 2.5 r = 3.0 r = 3.5

1 0.500 0.500 0.500 0.500 0.500 0.500 0.500
2 0.125 0.250 0.375 0.500 0.625 0.750 0.875
3 0.055 0.188 0.352 0.500 0.586 0.562 0.383
4 0.026 0.152 0.342 0.500 0.607 0.738 0.827
5 0.013 0.129 0.338 0.500 0.597 0.580 0.501
6 0.006 0.112 0.335 0.500 0.602 0.731 0.875
7 0.003 0.100 0.334 0.500 0.599 0.590 0.383
8 0.002 0.090 0.334 0.500 0.600 0.726 0.827
9 0.001 0.082 0.334 0.500 0.600 0.597 0.501
10 0.000 0.075 0.333 0.500 0.600 0.722 0.875
11 0.000 0.069 0.333 0.500 0.600 0.603 0.383
12 0.000 0.065 0.333 0.500 0.600 0.718 0.827
13 0.000 0.060 0.333 0.500 0.600 0.607 0.501
14 0.000 0.057 0.333 0.500 0.600 0.716 0.875
15 0.000 0.054 0.333 0.500 0.600 0.610 0.383
16 0.000 0.051 0.333 0.500 0.600 0.713 0.827
17 0.000 0.048 0.333 0.500 0.600 0.613 0.501
18 0.000 0.046 0.333 0.500 0.600 0.711 0.875
19 0.000 0.044 0.333 0.500 0.600 0.616 0.383
20 0.000 0.042 0.333 0.500 0.600 0.710 0.827

Figure 1 visualizes the resulting time series as a graph produced by Pynamical, with time on the
x-axis and the system state on the y-axis. This graph visualizes how the population changes over time
at different growth rates. For instance, the violet line for a growth rate of 0.5 quickly drops to zero:
the population dies out. The teal line that represents a growth rate of two (the replacement rate) stays
steady at a population level of 0.5. The growth rates of three and 3.5 are more interesting. While the
green line for a growth rate of three seems to slowly converge toward a stable value, the yellow line
for a growth rate of 3.5 just seems to repeatedly bounce around four different values.
Systems 2016, 4, 37 4 of 18

green line for a growth rate of three seems to slowly converge toward a stable value, the yellow line
for a growth
Systems rate of 3.5 just seems to repeatedly bounce around four different values.
2016, 4, 37 4 of 18
An attractor is the value, or set of values, that a system settles toward over time. When the
growth rate parameter is set to 0.5, the system has a fixed-point attractor at a population level of zero,
An attractor
as depicted by the is the
violetvalue,
line.orInset of values,
other words,thatthe apopulation
system settlesvaluetoward over toward
is drawn time. When the
a stable
growth rate parameter is set to 0.5, the system has a fixed-point attractor
equilibrium of zero over time as the model iterates: the logistic equation maps the value of a at a population level of zero,
as depicted by
fixed-point the violet
attractor to line.
itself.InWhen
other words, the population
the growth value isisdrawn
rate parameter set totoward
3.5, thea stable
systemequilibrium
oscillates
of zero over time as the model iterates: the logistic equation maps the value
between four values as depicted by the yellow line. This oscillating attractor is called a limit of a fixed-point attractor
cycle.
However, when we adjust the growth rate parameter beyond 3.57, we witness the onset of chaos. as
to itself. When the growth rate parameter is set to 3.5, the system oscillates between four values A
depictedsystem
chaotic by the yellow line. This
has a strange oscillating
attractor, attractor
around whichis called a limit cycle.
the system However,
oscillates forever when we adjust
without ever
the growthitself
repeating rate or
parameter
settling beyond 3.57, we
into a steady witness
state the onset
of behavior of chaos.
[45,46]. A chaotic
It never system
produces thehas a strange
same value
attractor, around which the system oscillates forever without ever repeating
twice, and its structure is fractal, meaning the same patterns exist at every scale no matter how much itself or settling into
a steady state of
we zoom into it [47]. behavior [45,46]. It never produces the same value twice, and its structure is fractal,
meaning the same patterns exist at every scale no matter how much we zoom into it [47].

Figure 1. Time series graph of the logistic map with seven growth rate parameter values over
Figure 1. Time series graph of the logistic map with seven growth rate parameter values over
20 generations.
20 generations.

3. System Bifurcations
3. System Bifurcations
To show this more clearly, we run the logistic model again, this time for 200 generations across
To show this more clearly, we run the logistic model again, this time for 200 generations across
1000 growth rate values between zero and four. When we produced the plot in Figure 1, we had only
1000 growth rate values between zero and four. When we produced the plot in Figure 1, we had only
seven growth rates. This time, we have 1000, so we need to visualize the results in a different way
seven growth rates. This time, we have 1000, so we need to visualize the results in a different way to
to make them comprehensible, using a bifurcation diagram that visualizes a system’s attractors as
make them comprehensible, using a bifurcation diagram that visualizes a system’s attractors as a
a function of some parameter [32,48,49]. The bifurcation diagram in Figure 2 represents 1000 discrete
function of some parameter [32,48,49]. The bifurcation diagram in Figure 2 represents 1000 discrete
vertical slices, each corresponding to one of 1000 growth rate parameter values evenly spaced between
vertical slices, each corresponding to one of 1000 growth rate parameter values evenly spaced
zero and four. To produce each of these visual slices, Pynamical ran the model 200 times, then threw
between zero and four. To produce each of these visual slices, Pynamical ran the model 200 times,
away the first 100 results, leaving just the final 100 generations for each growth rate. Each vertical slice
then threw away the first 100 results, leaving just the final 100 generations for each growth rate. Each
thus visualizes the population values that the logistic map settles toward over time (i.e., the attractor)
vertical slice thus visualizes the population values that the logistic map settles toward over time (i.e.,
for that parameter value.
the attractor) for that parameter value.
In Figure 2, we can see that for growth rates less than one, the system always eventually collapses
In Figure 2, we can see that for growth rates less than one, the system always eventually collapses
to zero (extinction). For growth rates between one and three, the system always settles into an exact,
to zero (extinction). For growth rates between one and three, the system always settles into an exact,
stable population level. For instance, in the vertical slice above a growth rate of 2.5, there is only
stable population level. For instance, in the vertical slice above a growth rate of 2.5, there is only one
one population value represented (0.6), and it corresponds precisely to where the line for a growth rate
population value represented (0.6), and it corresponds precisely to where the line for a growth rate
of 2.5 settles in Figure 1’s time graph. At this parameter value, the system’s attractor is a fixed point
of 2.5 settles in Figure 1’s time graph. At this parameter value, the system’s attractor is a fixed point
at 0.6. However, for some growth rates, such as 3.9, the plot in Figure 2 shows 100 different values;
Systems 2016, 4, 37 5 of 18
Systems 2016, 4, 37 5 of 18

at 0.6. However, for some growth rates, such as 3.9, the plot in Figure 2 shows 100 different values;
in other words, a different value value for
for each
each of
ofits
its100
100generations.
generations.Here,
Here,the
thesystem
systemnever
neversettles
settlesinto
intoa
afixed
fixedpoint
pointorora alimit
limitcycle.
cycle.

Figure 2.2. Bifurcation diagram of 100 generations of the logistic map for 1000 growth
growth rate
rate parameter
parameter
values between zero
zero and four.
four. The vertical slice above each growth rate depicts the system’s attractor
at that rate.

Why is
Why is this
this visualization called aa bifurcation
visualization called bifurcation diagram?
diagram? If If we
we zoom
zoom into
into the
the growth
growth ratesrates between
between
2.8 and 4 to see what is happening at a finer scale (Figure 3), the possible population values fork fork
2.8 and four to see what is happening at a finer scale (Figure 3), the possible population values into
into discrete
two two discrete
pathspaths
at theatvertical
the vertical slice above
slice above a growth
a growth rate ofrate
three.of At
three. At a growth
a growth ratethe
rate of 3.2, of 3.2,
systemthe
system oscillates
oscillates exclusively exclusively
betweenbetween two population
two population values: onevalues:
around one0.5around 0.5other
and the and around
the other around
0.8. Thus,
0.8. Thus, at that growth rate, applying the logistic map to one of these two population
at that growth rate, applying the logistic map to one of these two population values yields the other. values yields
the other.
Just beyondJust beyondrate
a growth a growth ratediagram
of 3.4, the of 3.4, the diagram
bifurcates bifurcates
again again
into four intoThis
paths. fourcorresponds
paths. This
corresponds
to the yellowtoline theinyellow
Figure line
1: in Figure
when the1:growth
when the rategrowth rate parameter
parameter is the
is set to 3.5, set to 3.5, the
system system
oscillates
over four population values. These are periods, just like the period of a pendulum. At a growth At
oscillates over four population values. These are periods, just like the period of a pendulum. ratea
of 3.2, the system has a period-two attractor. At a growth rate of 3.5, the system has a period-foura
growth rate of 3.2, the system has a period-two attractor. At a growth rate of 3.5, the system has
period-four
attractor. Justattractor.
beyond aJust beyond
growth rate aofgrowth rate of 3.5,
3.5, it bifurcates it bifurcates
again into eight again
paths into
as theeight
system paths as the
oscillates
system oscillates over eight population values. These consecutive bifurcations
over eight population values. These consecutive bifurcations are phase transitions from one behavior, are phase transitions
from as
such one behavior, such
a fixed-point as a fixed-point
attractor, attractor,
to a qualitatively to a qualitatively
different different
type of behavior, suchtypeas aof behavior, limit
period-two such
as a period-two limit cycle attractor, as we vary the parameter value. Beyond a
cycle attractor, as we vary the parameter value. Beyond a growth rate of 3.57, however, the bifurcations growth rate of 3.57,
however, the bifurcations ramp up until the system is capable of eventually
ramp up until the system is capable of eventually landing on any population value. This is known landing on any
population
as value. This is
the period-doubling known
path as theAs
to chaos. period-doubling
we adjust the growthpath torate
chaos. As we adjust
parameter upwards, the growth rate
the logistic
parameter upwards, the logistic map will oscillate between two, then four,
map will oscillate between two, then four, then eight, then 16, then 32 (and on and on to infinity)then eight, then 16, then
32 (and on and
population on to infinity) population values.
values.
Systems 2016, 4, 37 6 of 18

Systems 2016, 4,
Systems 2016, 4, 37
37 66 of
of 18
18

Figure 3. Bifurcation diagram of 100 generations of the logistic map for 1000 growth rate parameter
values 3.
Figure between 2.8 and
Bifurcation 4. The of
diagram vertical slice aboveofeach
100 generations growth map
the logistic rate depicts
for 1000the system’s
growth rateattractor
parameter at
Figure 3. Bifurcation diagram of 100 generations of the logistic map for 1000 growth rate parameter
that rate.
values between 2.8 and 4. The vertical slice above each growth rate depicts the system’s attractor at
values between 2.8 and 4. The vertical slice above each growth rate depicts the system’s attractor at
that rate.
that rate.
By the time we reach a growth rate of 3.99, it has bifurcated so many times that the system now
jumps,By seemingly randomly, between allof population values. We only say seemingly randomly because
By the
the time
time we
we reach
reach aa growth
growth raterate of 3.99,
3.99, it
it has
has bifurcated
bifurcated soso many
many times
times that the system
that the system now
now
it is definitely not truly random. Rather, this model follows very simple deterministic rules yet
jumps,
jumps, seemingly
seemingly randomly,
randomly,between
betweenall allpopulation
populationvalues.
values.WeWeonly
onlysaysayseemingly
seeminglyrandomly
randomly because
because it
produces
is definitely apparent
not trulyrandomness
random. due
Rather, to its
this attractor
model having
follows a
very period
simple of infinite
deterministic length.
rules This
yet is chaos:
produces
it is definitely not truly random. Rather, this model follows very simple deterministic rules yet
deterministic and aperiodic.
apparent If we zoom having
in again, to the narrow slice of growth rates between 3.7 and
producesrandomness due to its
apparent randomness attractor
due to its a period
attractor of infinite
having a periodlength. This is
of infinite chaos:
length. deterministic
This is chaos:
3.9 (Figure
and aperiodic.4), we begin
If we zoomto in
seeagain,
the visceral beauty slice
to theinnarrow of chaos. Out ofrates
the noise emerge strange swirling
deterministic and aperiodic. If we zoom again, to theofnarrowgrowth slice ofbetween
growth rates3.7 and 3.9 (Figure
between 4),
3.7 and
patterns
we begin and
to thresholds
see the on
visceral either
beauty side
of of
chaos.which
Out the
of system
the noise behaves
emerge very
strange differently.
swirling For example,
patterns and
3.9 (Figure 4), we begin to see the visceral beauty of chaos. Out of the noise emerge strange swirling
between theongrowth
thresholds either rates
sideonof which
of 3.82 and the3.84,
systemthe system moves from chaos back into order, oscillating
patterns and thresholds either side of whichbehaves verybehaves
the system differently.
veryFor example,
differently. between
For example, the
between
growth just three population values: approximately 0.15, 0.55 and 0.95. However, then at growth
betweenrates of 3.82 and
the growth rates3.84, the and
of 3.82 system
3.84,moves from chaos
the system movesbackfrominto order,
chaos backoscillating
into order,between just
oscillating
rates beyond
three population 3.86, it bifurcates
values: again and
approximately returns
0.15, 0.55 to chaos.
and 0.95. Indeed any
However, thenone-dimensional
at growth rates system
beyond with
3.86,
between just three population values: approximately 0.15, 0.55 and 0.95. However, then at growth
a bifurcates
it period-three cycle such as this at some parameter value is capable of chaotic behavior at other
rates beyondagain 3.86,and returns
it bifurcates toagain
chaos.andIndeed
returnsany toone-dimensional
chaos. Indeed any system with a period-three
one-dimensional system cycle
with
parameter
such values [50].
a period-three cycle such as this at some parameter value is capable of chaotic behavior at[50].
as this at some parameter value is capable of chaotic behavior at other parameter values other
parameter values [50].

Figure 4. Bifurcation diagram of 100 generations of the logistic map for 1000 growth rate parameter
Figure between
values 4. Bifurcation diagram
3.7 and 3.9. Theofsystem
100 generations
moves fromof order
the logistic map
to chaos forback
and 1000again
growth rategrowth
as the parameter
rate
values
is between 3.7 and 3.9. The system moves from order to chaos and back again as the growth rate
adjusted.
Figure 4. Bifurcation diagram of 100 generations of the logistic map for 1000 growth rate parameter
is adjusted.
values between 3.7 and 3.9. The system moves from order to chaos and back again as the growth rate
is adjusted.
Systems 2016, 4,
Systems 2016, 4, 37
37 77 of
of 18
18

Universality refers to the phenomenon that very different systems can exhibit very similar
Universality refers to the phenomenon that very different systems can exhibit very similar
behavior regardless of their underlying dynamics. It is commonly associated with Mitchell
behavior regardless of their underlying dynamics. It is commonly associated with Mitchell
Feigenbaum’s discovery that all systems that undergo this period-doubling path to chaos obey a
Feigenbaum’s discovery that all systems that undergo this period-doubling path to chaos obey
mathematical constant [51,52]: the distance between consecutive bifurcations along the horizontal axis
a mathematical constant [51,52]: the distance between consecutive bifurcations along the horizontal axis
shrinks by a factor that asymptotically approaches 4.669, now known as Feigenbaum’s constant [44].
shrinks by a factor that asymptotically approaches 4.669, now known as Feigenbaum’s constant [44].
Regardless of the system’s specific dynamics, the ratio of the bifurcations on its road to chaos always
Regardless of the system’s specific dynamics, the ratio of the bifurcations on its road to chaos always
obeys this constant.
obeys this constant.
4. Fractals and Strange Attractors
4. Fractals and Strange Attractors
There is also a deep and universal connection between chaos and fractals [37]. In Figure 4, the
There is also a deep and universal connection between chaos and fractals [37]. In Figure 4,
bifurcations around a growth rate of 3.85 may look familiar. If we zoom in to the center one
the bifurcations around a growth rate of 3.85 may look familiar. If we zoom in to the center one
(Figure 5), we incredibly see the same structure that we saw earlier at the macro-level. In fact, if we
(Figure 5), we incredibly see the same structure that we saw earlier at the macro-level. In fact, if we
keep zooming infinitely in to this visualization, we will continue seeing the same structure and
keep zooming infinitely in to this visualization, we will continue seeing the same structure and patterns
patterns at finer and finer scales, forever. How can this possibly be? We mentioned earlier that chaotic
at finer and finer scales, forever. How can this possibly be? We mentioned earlier that chaotic systems
systems have strange attractors and their structure can be characterized as fractal [53–55]. Fractals
have strange attractors and their structure can be characterized as fractal [53–55]. Fractals are shapes
are shapes that are self-similar, meaning they have the same structure at every scale [56–58]. As we
that are self-similar, meaning they have the same structure at every scale [56–58]. As we zoom in
zoom in on them, we find smaller copies of the larger macro-structure. The bifurcation diagram (and
on them, we find smaller copies of the larger macro-structure. The bifurcation diagram (and thus,
thus, the attractor) of the logistic map is a fractal: at the fine scale in Figure 5, we see a tiny reiteration
the attractor) of the logistic map is a fractal: at the fine scale in Figure 5, we see a tiny reiteration of
of the same bifurcations, chaos and limit cycles we saw in Figure 1’s visualization of the full range of
the same bifurcations, chaos and limit cycles we saw in Figure 1’s visualization of the full range of
growth rates.
growth rates.

Figure 5. Bifurcation diagram of 100 generations of the logistic map for 1000 growth rate parameter
Figure 5. Bifurcation diagram of 100 generations of the logistic map for 1000 growth rate parameter
values between 3.84 and 3.856. This is the same structure that we saw earlier at the macro-level in
values between 3.84 and 3.856. This is the same structure that we saw earlier at the macro-level in
Figure 3, because chaotic systems’ strange attractors are fractal.
Figure 3, because chaotic systems’ strange attractors are fractal.

Another way to visualize this nonlinear time series is with a phase diagram, diagram, using a method
called state-space
state-space reconstruction
reconstructionthrough
throughdelay-coordinate
delay-coordinateembedding
embedding [35].
[35]. Simply
Simply put,put,
thisthis plots
plots the
the system’s value at generation t + 1 on the y-axis versus its value at t on the x-axis [59],
system’s value at generation t + 1 on the y-axis versus its value at t on the x-axis [59], giving us another giving us
another visual window
visual window into the qualitative
into the qualitative behaviorbehavior of the The
of the system. system. The
clever cleverofinsight
insight of this
this phase phase
diagram
diagram
is that it is that it embeds
embeds one-dimensional
one-dimensional timedata
time series series dataour
from from our logistic
logistic map intomaptwo-dimensional
into two-dimensionalstate
state space: an imaginary space that uses system variables as its dimensions [33,60,61].
space: an imaginary space that uses system variables as its dimensions [33,60,61]. Each point in state Each point
in stateisspace
space is a state,
a system system or state, or in
in other otherawords,
words, a set of variable
set of variable values.
values. While While traditional
traditional systems’ systems’
analysis
analysis
tends to tends
focus toonfocus on visualizing
visualizing time series
time series as in Figure
as in Figure 1, nonlinear
1, nonlinear dynamics
dynamics tendstends to focus
to focus on
visualizing these state spaces. Few real-world systems are fully observable, yet the dynamics in a
properly reconstructed state space are identical to the true dynamics of the entire system [34].
Systems 2016, 4, 37 8 of 18

on visualizing
Systems 2016, 4, 37 these state spaces. Few real-world systems are fully observable, yet the dynamics 8 ofin
18
a properly reconstructed state space are identical to the true dynamics of the entire system [34].
In our
In our case,
case,the thetwo
twovariables
variablesare:
are:(1)
(1)the
thepopulation
populationvaluevalueatat generation
generation t; and
t; and (2)(2)
thethe value
value at
tat+t1.+ 1.
ForFor example,
example, witha agrowth
with growthrate
rateofof3.5,
3.5,the
thepopulation
populationvalue
valueatatGeneration
Generation11isis0.5;0.5; the
the value
value
at Generation
at Generation 22 isis 0.875;
0.875;the
thevalue
valueatatGeneration
Generation 3 is
3 0.383; andand
is 0.383; so forth (see(see
so forth Table 1). Therefore,
Table our
1). Therefore,
two-dimensional phase diagram will have (x, y) points at (0.5, 0.875) and (0.875,
our two-dimensional phase diagram will have (x, y) points at (0.5, 0.875) and (0.875, 0.383), and so on 0.383), and so on
(Figure 6B).
(Figure 6B). Remember
Remember that that our
our model
model follows
follows aa simple
simple deterministic
deterministic rule,
rule, so
so if
if we
we know
know aa certain
certain
generation’s population value, we can easily determine the next generation’s
generation’s population value, we can easily determine the next generation’s value. Like earlier, value. Like earlier, to
produce these phase diagrams, Pynamical runs the logistic model for
to produce these phase diagrams, Pynamical runs the logistic model for 200 generations and then200 generations and then
discards the
discards the first
first 100
100 rows,
rows, toto visualize
visualizeonly
onlythose
thosevalues
valuesthat
thatthe
thesystem
systemsettles
settlestoward
towardoverovertime.
time.

Figure
Figure 6.6. Phase
Phase diagrams
diagramsof ofthe
thelogistic
logisticmap
mapoverover200
200generations
generations forfor
growth rate
growth parameter
rate values
parameter of:
values
2.9 (A); 3.5 (B); 3.56 (C); and 3.57 (D). When the parameter is set to 2.9, the model converges
of: 2.9 (A); 3.5 (B); 3.56 (C); and 3.57 (D). When the parameter is set to 2.9, the model converges at a at a single
fixed-point. When the
single fixed-point. parameter
When is set to 3.5
the parameter or to
is set higher,
3.5 orthe modelthe
higher, oscillates
model over four points,
oscillates thenpoints,
over four eight,
and on and on as it bifurcates.
then eight, and on and on as it bifurcates.

In Figure 6A, the phase diagram shows that the logistic map homes in on a fixed-point attractor
at 0.655 (on both axes) when the growth rate parameter is set to 2.9. This corresponds to the vertical
slice above the x-axis value of 2.9 in the bifurcation diagram in Figure 2. Figure 6B depicts a period-
four limit cycle attractor: when the growth rate is set to 3.5, the logistic map oscillates over four points,
as shown in this phase diagram (and in Figures 1 and 2). If we adjust the growth rate parameter up
Systems 2016, 4, 37 9 of 18

In Figure 6A, the phase diagram shows that the logistic map homes in on a fixed-point attractor at
0.655 (on both axes) when the growth rate parameter is set to 2.9. This corresponds to the vertical slice
above the x-axis value of 2.9 in the bifurcation diagram in Figure 2. Figure 6B depicts a period-four
limit cycle
Systems 2016, attractor:
4, 37 when the growth rate is set to 3.5, the logistic map oscillates over four points, 9 of 18
as shown in this phase diagram (and in Figures 1 and 2). If we adjust the growth rate parameter up
to 3.56,
to 3.56, we
we witness
witness aa period-doubling
period-doubling bifurcation:
bifurcation: Figure
Figure 6C 6C shows
shows thethe system
system now
now oscillating
oscillating over
over
eight points.
eight points. As As wewe approach
approach thethe chaotic
chaotic regime,
regime, the
the range
range ofof parameter
parameter values
values inin which
which ourour system
system
behaves chaotically, the period-doubling bifurcations start to come more quickly.
behaves chaotically, the period-doubling bifurcations start to come more quickly. Figure 6D shows Figure 6D shows
that
that several
several additional
additional bifurcations
bifurcations occurred
occurred betweenbetween the growth
the growth rates rates ofand
of 3.56 3.563.57.
and 3.57.
AA kind
kind of of structure
structure is
is slowly
slowly being
being revealed
revealed across
across Figure
Figure 6,6, but
but we
we can
can see
see it
it much
much more
more clearly
clearly
as we push the growth rate parameter value deep into the chaotic regime.
as we push the growth rate parameter value deep into the chaotic regime. The phase diagram in The phase diagram in
Figure 7A
Figure 7A reveals
reveals thethe system’s
system’s attractor
attractor at
at aa growth
growth raterate of
of 3.9.
3.9. Figure
Figure 7B
7B visualizes
visualizes 5050 different
different
growth rate
growth rate parameter
parameter values
values between
between 3.63.6 and
and 4,4, each
each with
with its
its own
own color.
color. Those
Those rates
rates that
that exhibit
exhibit
chaos form parabolas, but some gaps exist where the system occasionally settles
chaos form parabolas, but some gaps exist where the system occasionally settles down into periodic down into periodic
behavior (e.g.,
behavior (e.g., in
in the
the teal
teal band
band when
when the
the growth
growth rate
rate is
is set
set to
to 3.83;
3.83; compare
compare this
this band
band of
of periodicity
periodicity
with Figure
with Figure 4). 4).

Figure 7. Cropped
Figure 7. Cropped phasephase diagrams
diagrams of of the
the logistic
logistic map
map over
over 200
200 generations
generations for:
for: (A)
(A) aa growth
growth rate
rate
parameter
parametervalue
valueofof3.9; and
3.9; (B)(B)
and 50 growth rate parameter
50 growth valuesvalues
rate parameter between 3.6 and3.6
between 4 (the
andchaotic
4 (theregime),
chaotic
each witheach
regime), its own
withcolored
its ownline
colored line

Strangeattractors
Strange attractorsare arerevealed
revealedby bythese
theseshapes
shapesasasthethesystem
systemisissomehow
somehow oddly
oddly constrained,
constrained, yetyet
it
it never settles into a fixed point or limit cycle like it did in Figure 6. Instead it
never settles into a fixed point or limit cycle like it did in Figure 6. Instead it just bounces around just bounces around
different population
different population values values (i.e.,
(i.e., points
points onon the
the parabola)
parabola) forever
forever without
without ever
ever repeating
repeating the the same
same
value twice.
value twice. ItIt is
is impossible
impossibleto topredict
predictififany
anytwotwoconsecutive
consecutiveobservations
observations appear
appear near
neareach other
each otheror
far apart on the parabola. Further, the parabolas in Figure 7B never overlap
or far apart on the parabola. Further, the parabolas in Figure 7B never overlap due to their fractal due to their fractal
geometryand
geometry andthethe deterministic
deterministicnature
natureof ofthe
the logistic
logisticmap.
map. Consider:
Consider: if if two
two different
different parameter
parameter values
values
could ever
could ever land
land on on the
the exact
exact same
same point,
point, their
their systems
systems would
would have
have toto evolve
evolve identically
identically over
over time
time
because the logistic map is deterministic. We can see in these visualizations
because the logistic map is deterministic. We can see in these visualizations that this indeed never that this indeed never
happens.While
happens. Whilethe thedynamics
dynamicsofofa chaotic
a chaotic system
system appear
appear to have
to have no pattern
no pattern whatsoever,
whatsoever, in reality,
in reality, they
they conform to a remarkable fractal pattern, a strange attractor, which confines
conform to a remarkable fractal pattern, a strange attractor, which confines the system to a limited slice the system to a
limited slice of state space and ensures that no state will ever repeat [62]. Fractals
of state space and ensures that no state will ever repeat [62]. Fractals are indeed strange. Rather than are indeed strange.
Rather than
having having a whole-number
a whole-number dimension such dimension
as twosuch as two
or three, or three,
they they are characterized
are characterized by a fractionalby a
fractional (hence, fractal) dimension [55,61,63]. The fractal dimension refers
(hence, fractal) dimension [55,61,63]. The fractal dimension refers to the space-filling characteristics ofto the space-filling
characteristics of a curve that, through self-similarity, becomes a bit more than a one-dimensional
line yet a bit less than a two-dimensional plane.
These visualizations have all plotted quantitative data to better explain and understand the
qualitative behavior of a nonlinear dynamical system. A cobweb plot is a visualization technique
particularly well-suited to revealing the qualitative behavior of one-dimensional maps, allowing us
to analyze the long-term evolution of such systems under recursive iteration [37]. The cobweb plots
Systems 2016, 4, 37 10 of 18

a curve that, through self-similarity, becomes a bit more than a one-dimensional line yet a bit less than
a two-dimensional plane.
These visualizations have all plotted quantitative data to better explain and understand the
qualitative behavior of a nonlinear dynamical system. A cobweb plot is a visualization technique
particularly well-suited to revealing the qualitative behavior of one-dimensional maps, allowing us
to analyze the long-term evolution of such systems under recursive iteration [37]. The cobweb plots
Systems
drawn by2016, 4, 37
Pynamical in Figure 8 consist of three lines: a diagonal gray identity line representing 10 ofy18
= x,
a red curve representing the logistic map as y = f(x) for a given parameter value and a blue line tracing
y = x, a red curve representing the logistic map as y = f(x) for a given parameter value and a blue line
the path of the cobweb. To draw a cobweb:
tracing the path of the cobweb. To draw a cobweb:
1. Begin on the x-axis at the point (x, 0) where x is the initial population value (0.5 in our example),
1. Begin on the x-axis at the point (x, 0) where x is the initial population value (0.5 in our example),
and draw a vertical line to the red function curve; this new point is at (x, f(x)).
and draw a vertical line to the red function curve; this new point is at (x, f(x)).
2. 2. Draw
Draw a ahorizontal
horizontalline
linefrom
fromthis
thispoint
point to
to the gray identity
the gray identityline;
line;this
thisnew
newpoint
pointis is
atat (f(x),
(f(x), f(x)).
f(x)).
3. 3. Draw a vertical line from this point to the red function curve; this new point is at (f(x),
Draw a vertical line from this point to the red function curve; this new point is at (f(x), f(f(x))). f(f(x))).
4. 4. Repeat
Repeat Steps
Steps2 2and
and33recursively.
recursively.The
The cobwebs Figure 88were
cobwebs in Figure wereiterated
iterated100
100times.
times.

Figure
Figure Cobweb
8. 8. Cobwebplots
plotsofofthe
thelogistic
logisticmap
map for
for growth rate parameter
growth rate parametervalues
valuesof:of:(A)
(A)
1; 1;
(B)(B) 2.7;
2.7; (C)(C)
3.5;3.5;
(D)(D)
3.9.3.9.The represents y = x;
The diagonal gray identity line represents y = x; the red curve represents the logistic map map
diagonal gray identity line the red curve represents the logistic as
as yy =
= f(x) for each
f(x) for each of
of the
the four
four parameter
parametervalues;
values;and
andthe
theblue
bluecobweb
cobwebline
linerepresents
represents the
the system’s
system’s
trajectory
trajectory over 100
over generations.
100 generations.

The blue lines intersect the red curve at those values our system lands on as it iterates from an
initial population value of 0.5. In Figure 8A,B, the cobweb shows the system homing in on fixed-point
attractors of zero and 0.65, respectively. At a growth rate of 3.5 (Figure 8C), the system oscillates over
four points in its limit cycle attractor, denoted by rectangular closed loops. The points where the blue
Systems 2016, 4, 37 11 of 18

The blue lines intersect the red curve at those values our system lands on as it iterates from
an initial population value of 0.5. In Figure 8A,B, the cobweb shows the system homing in on fixed-point
attractors of zero and 0.65, respectively. At a growth rate of 3.5 (Figure 8C), the system oscillates over
four points in its limit cycle attractor, denoted by rectangular closed loops. The points where
Systems 2016, 4, 37
the
11 of 18
blue lines intersect the red curve are the same as those revealed by the attractor in Figure 6B for the
same
rate of parameter
3.9. Thevalue.chaoticFinally,
orbit Figure
fills the8Dplot
visualizes our system’sanbehavior
with rectangles, in the
eventually chaoticnumber
infinite regime at of
anever-repeating
growth rate of trajectories
3.9. The chaotic orbit fills the plot with rectangles,
that form a fractal cobweb throughout the diagram. an eventually infinite number of
never-repeating trajectories that form a fractal cobweb throughout the diagram.
5. Chaos and Randomness
5. Chaos and Randomness
Phase diagrams are useful for visually revealing strange attractors in time series data, like that
Phase
produced by diagrams are useful
the logistic for visually
map, because theyrevealing
embed this strange attractors in time
one-dimensional data series
into adata,
two- like that
or even
produced by the logistic
three-dimensional map,Itbecause
state space. they embed
can be difficult this one-dimensional
to ascertain data into
if certain time series a two- or even
are deterministic or
three-dimensional
just random if we do not fully understand their underlying dynamics [64]. Take the two deterministic
state space. It can be difficult to ascertain if certain time series are series plotted
or
by just randominifFigure
Pynamical we do9not fully
as an understand
example. Both their
of theunderlying
lines seem dynamics [64]. Take
to jump around the two
randomly. series
The red
plotted
line does bydepict
Pynamical
random in data,
Figurebut9 asthean example.
blue line comesBothfrom
of the ourlines seem
logistic to jump
model whenaround randomly.
the growth rate
The
is setred line does
to 3.99. This isdepict random data,
deterministic chaos,butbuttheit isblue line comes
difficult from ourfrom
to differentiate logistic model when
randomness. the
Instead
growth
in Figure rate10,is we
set to 3.99. This
visualize is deterministic
these same two datasetschaos, butwithit isphase
difficult to differentiate
diagrams fromtime
rather than randomness.
graphs,
Instead in Figure 10, we visualize these same two datasets with
giving us a clear window into the qualitative behavior of our systems. Now, we can clearly phase diagrams rather than
seetime
our
graphs,
chaotic system constrained by its strange attractor. By contrast, the random data just look like see
giving us a clear window into the qualitative behavior of our systems. Now, we can clearly the
our
noisechaotic
that theysystem constrained
actually are. by its strange attractor. By contrast, the random data just look like the
noise that they actually are.

Figure 9. Plot of two time series, one deterministic/chaotic from the logistic map (blue), and one
Figure 9. Plot of two time series, one deterministic/chaotic from the logistic map (blue), and one
random (red).
random (red).

This
This is particularly revealing
is particularly revealing in
in aa three-dimensional
three-dimensional phase
phase diagram
diagram from
from Pynamical
Pynamical (Figure
(Figure 10B)
10B)
that embeds our time series into a three-dimensional state space by plotting the population
that embeds our time series into a three-dimensional state space by plotting the population value at value
at generation
generation t +t 2+versus
2 versus
thethe
value at tat+ t1+versus
value 1 versus
the the value
value t. This
at t.atThis plotplot essentially
essentially extrudes
extrudes our
our two-dimensional plot (Figure 10A), then pans and rotates the viewpoint. In fact,
two-dimensional plot (Figure 10A), then pans and rotates the viewpoint. In fact, if we looked straight if we looked
straight
down atdown at the xy-plane
the xy-plane of the three-dimensional
of the three-dimensional plot in plot in Figure
Figure 10B, it 10B,
wouldit would look identical
look identical to the
to the two-dimensional plot in Figure 10A (see Appendix A for an animated visualization
two-dimensional plot in Figure 10A (see Appendix A for an animated visualization of this). Strange of this).
attractors stretch and fold state space in higher dimensions, allowing their fractal forms to fill space
without ever producing the same value twice.
Systems 2016, 4, 37 12 of 18

Strange attractors stretch and fold state space in higher dimensions, allowing their fractal forms to fill
space
Systemswithout
Systems2016,
2016,4,4,37ever producing the same value twice.
37 12
12of
of18
18

Figure
Figure 10. Phase
10.Phase
Figure10. diagrams
Phasediagrams of
diagramsof the
ofthe time
thetime series
timeseries in
seriesin Figure
Figure9.9.9.(A)
inFigure (A)
(A)isisisaaatwo-dimensional
two-dimensional state
two-dimensionalstate space
statespace version
spaceversion
version
(the
(the xy-plane)
xy-plane) of
of the
the three-dimensional
three-dimensional one
one
(the xy-plane) of the three-dimensional one (B). (B).
(B).

To
To press
Topress this
pressthis further,
thisfurther, we
further,we can
wecan use
canuse Pynamical
usePynamical
Pynamicalto to visualize
tovisualize
visualizethethe rest
therest of
restof the
ofthe logistic
thelogistic map’s
logisticmap’s chaotic
map’schaotic
chaotic
regime in three
regimeininthree
regime dimensions:
threedimensions:
dimensions:the the phase
thephase diagram
phasediagram in Figure
diagramininFigure
Figure1111 is a three-dimensional
11isisaathree-dimensional
three-dimensionalversion version
versionof of the
ofthe
the
two-dimensional
two-dimensionalstate
two-dimensional state space
statespace
spacewe we saw
wesaw in Figure
sawininFigure 7B.
Figure7B. The
7B.The novel
Thenovel color
novelcolor coding
colorcoding exposes
codingexposes
exposesthethe dynamical
thedynamical
dynamical
system’s
system’s behavior
system’sbehavior
behavior across
acrossacross the
the chaotic
the chaotic regime: regime:
chaotic regime: information
information information virtually
virtually impenetrable
virtually impenetrable impenetrable without
without
without visualization.
visualization.
visualization.
The beautiful TheThe beautiful
beautiful
structure structure
structure
of the strangeof the
of the strange attractor
strangeisattractor
attractor revealedis revealed
is revealed as it twists
as it twists
as it twists and
and curls curls
and curls
aroundaround
around
its
its
itsthree-dimensional
three-dimensional state
statespace
space (see
(see Appendix
Appendix AAfor
for an
an animated
animated visualization).
visualization).
three-dimensional state space (see Appendix A for an animated visualization). This structure again This
This structure
structure again
again
demonstrates
demonstratesthat
demonstrates that our
thatour apparently
ourapparently random
apparentlyrandom
randomtimetime series
timeseries data
seriesdata from
datafrom
fromthethe logistic
thelogistic model
modelisisisnot
logisticmodel not truly
nottruly random
trulyrandom
random
at all.
atatall. Instead,
all.Instead, it is aperiodic
Instead,ititisisaperiodic deterministic
aperiodicdeterministic chaos,
deterministicchaos, constrained
chaos,constrained
constrainedby by a mind-bending
byaamind-bending
mind-bendingstrange strange attractor.
strangeattractor.
attractor.
No
No matter
matter how
how much
much we
we zoom
zoom in,
in, the
the parabolas
parabolas never
never overlap
overlap and
and
No matter how much we zoom in, the parabolas never overlap and no point ever repeats itself. no
no point
point ever
ever repeats
repeats itself.
itself.

Figure 11. Two different viewing perspectives of a single three-dimensional phase diagram of the
Figure
Figure11.
11.Two
Two different
different viewing
viewingperspectives
perspectivesof
of aasingle
singlethree-dimensional
three-dimensional phase
phasediagram
diagramof of the
the
logistic map over 200 generations for 50 growth rate parameter values between 3.6 and 4, each with its
logistic map over 200 generations for 50 growth rate parameter values between 3.6 and 4, each
logistic map over 200 generations for 50 growth rate parameter values between 3.6 and 4, each with with
own colored line.
its
itsown
owncolored
coloredline.
line.
Systems 2016, 4, 37 13 of 18
Systems 2016, 4, 37 13 of 18

6. Unpredictable
UnpredictableSystems:
Systems:The
TheButterfly
Butterfly Effect
Effect
Attractors
Attractors have
have aa basin
basin of
of attraction:
attraction: a set of points that the system’s dynamics will pull into this
attractor
attractor over
over time
time [65].
[65]. This is easily seen with a cobweb plot. Figure 12 shows how the logistic map’s
basin
basin of attraction (when the growth rate is 2.7) pulls three different different initial population
population values
values into the
same
same fixed-point
fixed-pointattractor.
attractor.The initial
The statestate
initial of theofsystem
the systemwill eventually becomebecome
will eventually unknowable, because
unknowable,
any
because any one of many different possible points in the basin of attraction could have been the into
one of many different possible points in the basin of attraction could have been the one pulled one
the attractor.
pulled into the attractor.
By contrast,
contrast,chaotic
chaoticsystems
systemsare characterized
are characterized by their bysensitive dependence
their sensitive on initial conditions.
dependence on initial
Their strange attractors are globally stable, yet locally unstable: they
conditions. Their strange attractors are globally stable, yet locally unstable: they have basins of have basins of attraction,
yet within ayet
attraction, strange
withinattractor,
a strangeinfinitesimally close pointsclose
attractor, infinitesimally diverge overdiverge
points time without
over timeeverwithout
leavingeverthe
attractor’s
leaving theconfines.
attractor’sThis divergence
confines. Thiscan be measured
divergence can be by measured
Lyapunovby exponents
Lyapunov [66], the calculation
exponents of
[66], the
which is described
calculation of which by is
Wolf et al. [67].
described byIfWolf
the Lyapunov
et al. [67].exponent’s value isexponent’s
If the Lyapunov positive, then the is
value two points
positive,
move
then theapart
twoover time move
points at an exponential
apart over rate.timeIfatthe anLyapunov
exponential exponent
rate. Ifisthe
negative,
Lyapunovthen these points
exponent is
converge exponentially quickly, such as toward a fixed point or limit
negative, then these points converge exponentially quickly, such as toward a fixed point or limitcycle. Finally, the Lyapunov
exponent is zero
cycle. Finally, thewhen there isexponent
Lyapunov a bifurcation
is zero[68].
whenForthere
example, with our logistic
is a bifurcation model,
[68]. For the Lyapunov
example, with our
exponent is zerothe
logistic model, when the growth
Lyapunov rate isisset
exponent to one
zero when or the
three because
growth they
rate areto
is set bifurcation
one or three points; it is
because
negative for most growth rates, such as 0 ≤ r < 1 and 1 < r < 3, because they have
they are bifurcation points; it is negative for most growth rates, such as 0 ≤ r < 1 and 1 < r < 3, because fixed-point or limit
cycle attractors;
they have and it is
fixed-point orpositive for the
limit cycle chaotic and
attractors; regime it is(exclusive
positive for of those occasional
the chaotic regime windows when
(exclusive of
the
thosesystem resumes
occasional brief periodicity,
windows when the systemsuch as when the
resumes briefgrowth rate issuch
periodicity, 3.83).
as A positive
when Lyapunov
the growth rate
exponent
is 3.83). A indicates that the system
positive Lyapunov exponenthas indicates
a highly sensitive
that the systemdependence on initial
has a highly conditions
sensitive and is
dependence
aoncommon signature and
initial conditions of chaos [69–71]. signature of chaos [69–71].
is a common

12. Cobweb plots of the logistic map pulling initial population values of 0.1 (A), 0.5 (B) and
Figure 12. and
0.9 (C) into the
the same
samefixed-point
fixed-pointattractor
attractorover
overtime.
time.AtAt
thisthis
growth raterate
growth parameter value
parameter of 2.7,
value the
of 2.7,
Lyapunov
the Lyapunovexponent is negative.
exponent is negative.

This nonlinear
This nonlinear divergence
divergence of of very
very similar
similar values
values makes
makes real-world
real-world modeling
modeling and and prediction
prediction
difficult, because we must measure the parameters and system state with infinite
difficult, because we must measure the parameters and system state with infinite precision. Otherwise, precision.
Otherwise, tiny errors in measurement or rounding are compounded over
tiny errors in measurement or rounding are compounded over time until the system eventually time until the system
eventually
diverges diverges drastically
drastically from the prediction.
from the prediction. In the realInworld,
the realinfinite
world,precision
infinite precision is impossible.
is impossible. It was
It was through one such rounding error that Lorenz first discovered chaos.
through one such rounding error that Lorenz first discovered chaos. Recall his words at the Recall his words at the
beginning
beginning of this article: “the present determines the future, but the approximate present
of this article: “the present determines the future, but the approximate present does not approximately does not
approximately
determine determine
the future” [39].the future” [39].
Systems 2016, 4, 37 14 of 18
Systems 2016, 4, 37 14 of 18

As a demonstration
As a demonstration of this, we
of this, we run
run the
the logistic
logistic model
model with
with two
two very
very similar
similar initial
initial population
population
values, shown in Figure 13. Both have the same growth rate parameter value of 3.9. The blue
values, shown in Figure 13. Both have the same growth rate parameter value of 3.9. The blue line
line
represents an initial population value of 0.5, and the red line represents an initial population
represents an initial population value of 0.5, and the red line represents an initial population of 0.50001. of
0.50001. These two initial conditions are extremely close to one another, and accordingly,
These two initial conditions are extremely close to one another, and accordingly, their trajectories look their
trajectories look essentially
essentially identical identical
for the first for the first
30 generations. 30 generations.
After that, however,After
the that, however,
minuscule the minuscule
difference in initial
difference in initial conditions compounds to the point that by the 40th generation,
conditions compounds to the point that by the 40th generation, the two lines show little in common.the two lines show
little in common. What began as nearly indistinguishable initial conditions produces
What began as nearly indistinguishable initial conditions produces completely different outcomes over completely
different
time due outcomes over time
to nonlinearity due to nonlinearity
and exponential and exponential divergence.
divergence.

Figure 13.
Figure Plotofoftwo
13. Plot two time
time series
series withwith identical
identical dynamics,
dynamics, one starting
one starting at an at an initial
initial population
population value
value of 0.5 (blue) and the other starting at 0.50001 (red). At this growth rate parameter value
of 0.5 (blue) and the other starting at 0.50001 (red). At this growth rate parameter value of 3.9, the of
3.9, the Lyapunov exponent is positive; thus, the system is chaotic, and we can see the nearby points
Lyapunov exponent is positive; thus, the system is chaotic, and we can see the nearby points diverge
diverge over time.
over time.

If our knowledge of these two systems began at Generation 50, we would have no way of
that they
guessing that theywere
werenearly
nearlyidentical
identicalinin
thethe beginning.
beginning. With
With chaos,
chaos, history
history is thus
is thus lost
lost to to time,
time, and
and prediction
prediction of theoffuture
the future
is onlyis only as accurate
as accurate as ourasmeasurements.
our measurements. Human Human measurements
measurements are
are never
never infinitely
infinitely precise,precise, so in real-world
so in real-world chaoticchaotic
systems,systems, errors compound,
errors compound, and the and
futurethebecomes
future becomes
entirely
entirely unknowable
unknowable given longgiven long time
enough enough time horizons.
horizons. This phenomenon
This phenomenon is famouslyis famously
known asknown as the
the butterfly
butterfly
effect: effect: a flaps
a butterfly butterfly flaps in
its wings itsChina
wingsandin China and
sets off sets off in
a tornado a tornado in Texas.
Texas. Small events Small events
compound
compound
and and alter
irreversibly irreversibly
the futurealter the universe.
of the future ofInthe universe.
Figure In Figure
13, a tiny 13, aoftiny
fluctuation fluctuation
0.00001 makes an of
0.00001 makes
enormous an enormous
difference differenceand
in the behavior in the behavior
state of the and
systemstate
40ofgenerations
the system later.
40 generations
Although later.
this
Although this system’s
system’s future cannot future cannot be
be predicted, wepredicted, we can characterize
can characterize its dynamicsits dynamics geometrically
geometrically with phase
with phase bifurcation
diagrams, diagrams, bifurcation
plots and plots and plots;
cobweb cobweband plots; and statistically
statistically with Lyapunov
with Lyapunov exponents
exponents and
and fractal
fractal dimensions.
dimensions.

7. Conclusions
7. Conclusions
This article
This article had
had three
threeaims.
aims.First,
First,ititargued
arguedfor
forthe
theusefulness
usefulness ofof several
several visualization
visualization methods
methods to
to analyze and understand the behavior of nonlinear dynamical systems. Second,
analyze and understand the behavior of nonlinear dynamical systems. Second, it introduced the it introduced the
foundational concepts
foundational conceptsof of
nonlinear dynamics,
nonlinear dynamics,chaos, fractals,
chaos, self-similarity
fractals, and the limits
self-similarity and theof prediction.
limits of
It argued that information visualization is a key way to engage with these system
prediction. It argued that information visualization is a key way to engage with these system concepts. Third,
concepts. Third, and by way of the first two, it demonstrated the Pynamical software package for
visualizing the behavior of discrete nonlinear dynamical systems. In particular, Pynamical translates
Systems 2016, 4, 37 15 of 18

and by way of the first two, it demonstrated the Pynamical software package for visualizing the
behavior of discrete nonlinear dynamical systems. In particular, Pynamical translates these concepts
and tools into the Python programming language because it is prevalent, fast and easy to use (details in
Appendix A). This package fills a need for a free, fast, simple, extensible tool to introduce and analyze
nonlinear dynamical systems’ behavior visually, making it a useful platform for research, engineering
and pedagogy.
Nonlinear systems are extremely difficult to solve analytically because they cannot be broken
down into constituent parts. Instead of solving them, this article presented visual geometric methods
to explore system behavior and stability. These methods have broad applicability to scholarly and
professional visual data analysis by revealing hidden structure and patterns in time series whose
underlying dynamics may not be well known. In particular, they reveal the qualitative behavior of
nonlinear dynamical systems over time and/or in response to parameter variations.
This article used the logistic map to define such a set of nonlinear dynamics. As simple as this
model was, at different growth rate parameter values, it produced stability, periodic oscillations
or chaos. We used Pynamical to create bifurcation diagrams and cobweb plots to visualize this
behavior across different parameter values. In the chaotic regime, the system jumped seemingly
randomly between all population values. Accordingly, we used Pynamical to embed the data into
higher-dimensional state space to create phase diagrams to visualize the system’s strange attractor
and understand its constrained, deterministic dynamics. Finally, we explored the butterfly effect’s
implications of nonlinearity on system sensitivity, as infinitesimal differences in initial conditions
compounded over time until nearly identical systems had diverged drastically. Thus, in many nonlinear
systems, there are fundamental limits to knowledge and prediction. These visualization methods
can help researchers discover, examine and understand the behavior of nonlinear dynamical systems
and chaos.

Conflicts of Interest: The authors declare no conflict of interest.

Appendix A
Pynamical and all of the code used to develop these models and produce these visualizations are
available in a public repository on GitHub at https://github.com/gboeing/pynamical. Pynamical is
built on top of Python’s numpy, numba, pandas and matplotlib code libraries:

• numpy is a numerical library that handles the underlying data vectors;

• numba provides just-in-time compilation for optimized performance;
• pandas handles the higher-level data structures and analysis;
• matplotlib is the engine used to produce the visualizations and graphics.

Python was selected for developing Pynamical and the visualizations in this article because it is
a standard programming environment for information visualization. It offers three key advantages:
prevalence, speed and ease of use. First, Python has become a widely-prevalent language. The TIOBE
index, a common measure of programming language popularity, ranks Python as the fifth most
popular programming language as of 2016. In contrast, other languages commonly used for data
science, visualization and mathematics are ranked much lower (e.g., R, MATLAB and SAS are only
ranked 18th, 19th and 21st, respectively, in popularity). Developing tools for exploring, understanding
and visualizing dynamical systems in Python makes them available to a much wider audience of
systems analysts, researchers and students. Furthermore, unlike many languages, Python is completely
free and open-source. It is also multi-purpose, and researchers can use its standard syntax and grammar
for everything from statistical modeling, to cartography, to full software development. Becoming
a Swiss army knife of the computational science world, Python has grown popular and powerful.
Today, innumerable researchers and developers contribute open-source libraries of pre-packaged
functionality for all to repurpose.
Systems 2016, 4, 37 16 of 18

Second, Pynamical is fast. It is built on top of Python’s pandas, numpy and numba libraries for
quick vectorized numerical computation. Although it is an interpreted language, these libraries provide
compiled components that make mathematical modeling fast and efficient and can take advantage of
the optimized mathematical routines of Intel processors’ Math Kernel Library. Pynamical also takes
advantage of just-in-time code compilation to offer additional speed-ups. Third, Python is very easy to
use, and this simplicity translates to Pynamical. Most of the visualizations in this article can be created
from scratch in just one or two lines of straightforward code. The following examples demonstrate this
simplicity. The user merely imports Pynamical into the Python environment, then runs the following
code to produce the visualization:
• Figure 2: bifurcation_plot(simulate(num_rates = 1000))
• Figure 4: bifurcation_plot(simulate(rate_min = 3.7, rate_max = 3.9, num_rates = 1000))
• Figure 6D: phase_diagram(simulate(num_gens = 100, rate_min = 3.57))
• Figure 8D: cobweb_plot(r = 3.9, x0 = 0.5)
• Figure 11: phase_diagram_3d(simulate(num_gens = 4000, rate_min = 3.6, num_rates = 50)).
Pynamical defines extensible functions to express a discrete map’s equation and encapsulate
a model to simulate it iteratively. The logistic map, the Singer map and the cubic map are built-in by
default, but any other iterated map can be defined and added by the user. Pynamical also defines
a function to convert model output into x-y points, as well as functions to plot these points as
a bifurcation diagram, a cobweb plot, an animated cobweb plot, a two-dimensional phase diagram,
a three-dimensional phase diagram and an animated three-dimensional phase diagram. Animated
cobweb plots of the entire parameter space and animated three-dimensional phase diagrams extending
those presented in this study are also available in the GitHub repository. They shed particular
light on the fractal nature of strange attractors as they stretch and fold state space, thus serving
as an indispensable tool for pedagogy and visual information presentation.

References
1. Hastings, A.; Hom, C.L.; Ellner, S.; Turchin, P.; Godfray, H.C.J. Chaos in Ecology: Is Mother Nature a Strange
Attractor? Annu. Rev. Ecol. Syst. 1993, 24, 1–33. [CrossRef]
2. Rickles, D.; Hawe, P.; Shiell, A. A Simple Guide to Chaos and Complexity. J. Epidemiol. Commun. Health 2007,
61, 933–937. [CrossRef] [PubMed]
3. Suetani, H.; Soejima, K.; Matsuoka, R.; Parlitz, U.; Hata, H. Manifold Learning Approach for Chaos in the
Dripping Faucet. Phys. Rev. E 2012, 86, 036209. [CrossRef] [PubMed]
4. Singh, S.L.; Mishra, S.N.; Sinkala, W. A New Iterative Approach to Fractal Models. Commun. Nonlinear Sci.
Numer. Simul. 2012, 17, 521–529. [CrossRef]
5. Hoshi, R.A.; Pastre, C.M.; Vanderlei, L.C.M.; Godoy, M.F. Poincaré Plot Indexes of Heart Rate Variability:
Relationships with Other Nonlinear Variables. Auton. Neurosci. 2013, 177, 271–274. [CrossRef] [PubMed]
6. Babbs, C.F. Initiation of Ventricular Fibrillation by a Single Ectopic Beat in Three Dimensional Numerical
Models of Ischemic Heart Disease: Abrupt Transition to Chaos. J. Clin. Exp. Cardiol. 2014, 5, 2–11. [CrossRef]
7. Glass, L. Introduction to Controversial Topics in Nonlinear Science: Is the Normal Heart Rate Chaotic? Chaos
2009, 19, 028501. [CrossRef] [PubMed]
8. Hong, Z.; Dong, J. Chaos Theory and Its Application in Modern Cryptography. In Proceedings of the
2010 International Conference on Computer Application and System Modeling (ICCASM 2010), Taiyuan,
China, 22–24 October 2010; pp. 332–334.
9. Makris, G.; Antoniou, I. Cryptography with Chaos. In Proceedings of the 5th Chaotic Modeling and
Simulation International Conference, Athens, Greece, 12–15 June 2012; pp. 309–318.
10. Guastello, S.J. Chaos, Catastrophe, and Human Affairs: Applications of Nonlinear Dynamics to Work, Organizations,
and Social Evolution; Psychology Press: New York, NY, USA, 2013.
11. Richards, D. From Individuals to Groups: The Aggregation of Votes and Chaotic Dynamics. In Chaos Theory
in the Social Sciences; Kiel, L.D., Elliott, E., Eds.; University of Michigan Press: Ann Arbor, MI, USA, 1996;
pp. 89–116.
Systems 2016, 4, 37 17 of 18

12. Batty, M.; Longley, P. Fractal Cities: A Geometry of Form and Function; Academic Press: London, UK, 1994.
13. Batty, M.; Xie, Y. Self-Organized Criticality and Urban Development. Discret. Dyn. Nat. Soc. 1999, 3, 109–124.
[CrossRef]
14. Benguigui, L.; Czamanski, D.; Marinov, M.; Portugali, Y. When and Where is a City Fractal? Environ. Plan. B
2000, 27, 507–519. [CrossRef]
15. Shen, G. Fractal Dimension and Fractal Growth of Urbanized Areas. Int. J. Geogr. Inf. Sci. 2002, 16, 419–437.
[CrossRef]
16. Chen, Y.; Zhou, Y. Scaling Laws and Indications of Self-Organized Criticality in Urban Systems. Chaos Solitons
Fractals 2008, 35, 85–98. [CrossRef]
17. Chen, W.C. Nonlinear Dynamics and Chaos in a Fractional-Order Financial System. Chaos Solitons Fractals
2008, 36, 1305–314. [CrossRef]
18. Guégan, D. Chaos in Economics and Finance. Annu. Rev. Control 2009, 33, 89–93. [CrossRef]
19. Puu, T. Attractors, Bifurcations, & Chaos: Nonlinear Phenomena in Economics, 2nd ed.; Springer Science &
Business Media: New York, NY, USA, 2013.
20. Rosser, J.B. Chaos Theory and Rationality in Economics. In Chaos Theory in the Social Sciences; Kiel, L.D.,
Elliott, E., Eds.; University of Michigan Press: Ann Arbor, MI, USA, 1996; pp. 199–213.
21. Oxley, L.; George, D.A.R. Economics on the Edge of Chaos: Some Pitfalls of Linearizing Complex Systems.
Environ. Model. Softw. 2007, 22, 580–589. [CrossRef]
22. Hamouche, M.B. Can Chaos Theory Explain Complexity In Urban Fabric? Applications in Traditional
Muslim Settlements. Nexus Netw. J. 2009, 11, 217–242. [CrossRef]
23. Ostwald, M.J. The Fractal Analysis of Architecture: Calibrating the Box-Counting Method Using Scaling
Coefficient and Grid Disposition Variables. Environ. Plan B 2013, 40, 644–663. [CrossRef]
24. Cartwright, T.J. Planning and Chaos Theory. J. Am. Plan. Assoc. 1991, 57, 44–56. [CrossRef]
25. Innes, J.E.; Booher, D.E. Planning with Complexity; Routledge: London, UK, 2010.
26. Batty, M.; Marshall, S. The Origins of Complexity Theory in Cities and Planning. In Complexity Theories of
Cities Have Come of Age; Portugali, J., Meyer, H., Stolk, E., Tan, E., Eds.; Springer: Berlin, Germany, 2012;
pp. 21–45.
27. Batty, M. The New Science of Cities; MIT Press: Cambridge, MA, USA, 2013.
28. Alpigini, J.J. Dynamical System Visualization and Analysis via Performance Maps. Inf. Vis. 2004, 3, 271–287.
[CrossRef]
29. Layek, G.C. An Introduction to Dynamical Systems and Chaos; Springer: New Delhi, India, 2015.
30. Chen, C. Information Visualization: Beyond the Horizon, 2nd ed.; Springer: London, UK, 2006.
31. Lorenz, E.N. Deterministic Nonperiodic Flow. J. Atmos. Sci. 1963, 20, 130–141. [CrossRef]
32. May, R.M. Simple Mathematical Models with Very Complicated Dynamics. Nature 1976, 261, 459–467.
[CrossRef] [PubMed]
33. Packard, N.H.; Crutchfield, J.P.; Farmer, J.D.; Shaw, R.S. Geometry from a Time Series. Phys. Rev. Lett. 1980,
45, 712–716. [CrossRef]
34. Bradley, E. Time Series Analysis. In Intelligent Data Analysis: An Introduction, 2nd ed.; Hand, D., Berthold, M., Eds.;
Springer: Berlin, Germany, 2003.
35. Bradley, E.; Kantz, H. Nonlinear Time-Series Analysis Revisited. Chaos 2015, 25, 097610. [CrossRef] [PubMed]
36. Chettiparamb, A. Metaphors in Complexity Theory and Planning. Plan Theory 2006, 5, 71–91. [CrossRef]
37. Tomida, A.G. Matlab Toolbox and GUI for Analyzing One-Dimensional Chaotic Maps. In Proceedings of the
2008 International Conference on Computational Sciences and Its Applications, Perugia, Italy, 30 June–3 July 2008;
pp. 321–330.
38. Stewart, I. The Lorenz Attractor Exists. Nature 2000, 406, 948–949. [CrossRef] [PubMed]
39. Danforth, C.M. Chaos in an Atmosphere Hanging on a Wall. Available online: http://mpe2013.org/2013/
03/17/chaos-in-an-atmosphere-hanging-on-a-wall/ (accessed on 1 September 2016).
40. Gleick, J. Chaos: Making a New Science; Cardinal: Indianapolis, IN, USA, 1991.
41. May, R.M. Biological Populations with Nonoverlapping Generations: Stable Points, Stable Cycles, and Chaos.
Science 1974, 186, 645–647. [CrossRef] [PubMed]
42. Li, W.; Wang, K.; Su, H. Optimal Harvesting Policy for Stochastic Logistic Population Model. Appl. Math. Comput.
2011, 218, 157–162. [CrossRef]
Systems 2016, 4, 37 18 of 18

43. Pastijn, H. Chaotic Growth with the Logistic Model of P.F. Verhulst. In The Logistic Map and the Route to Chaos;
Ausloos, M., Dirickx, M., Eds.; Springer: Berlin, Germany, 2006; pp. 3–11.
44. Strogatz, S.H. Nonlinear Dynamics and Chaos, 2nd ed.; Westview Press: Boulder, CO, USA, 2014.
45. Ruelle, D.; Takens, F. On the Nature of Turbulence. Commun. Math. Phys. 1971, 20, 167–192. [CrossRef]
46. Shilnikov, L. Bifurcations and Strange Attractors. Proc. Int. Congr. Math. 2002, 3, 349–372.
47. Grebogi, C.; Ott, E.; Yorke, J.A. Chaos, Strange Attractors, and Fractal Basin Boundaries in Nonlinear
Dynamics. Science 1987, 238, 632–638. [CrossRef] [PubMed]
48. Gershenson, C. Introduction to Chaos in Deterministic Systems. Available online: http://arxiv.org/abs/
nlin/0308023 (accessed on 10 November 2016).
49. Wu, G.C.; Baleanu, D. Discrete Fractional Logistic Map and Its Chaos. Nonlinear Dyn. 2014, 75, 283–287.
[CrossRef]
50. Li, T.Y.; Yorke, J.A. Period Three Implies Chaos. Am. Math. Mon. 1975, 82, 985–992. [CrossRef]
51. Feigenbaum, M.J. Quantitative Universality for a Class of Nonlinear Transformations. J. Stat. Phys. 1978, 19,
25–52. [CrossRef]
52. Feigenbaum, M.J. Universal Behavior in Nonlinear Systems. Phys. Nonlinear Phenom. 1983, 7, 16–39. [CrossRef]
53. Hénon, M. A Two-Dimensional Mapping with a Strange Attractor. Commun. Math. Phys. 1976, 50, 69–77.
[CrossRef]
54. Farmer, J.D.; Ott, E.; Yorke, J.A. The Dimension of Chaotic Attractors. Phys. Nonlinear Phenom. 1983, 7, 153–180.
[CrossRef]
55. Grassberger, P.; Procaccia, I. Characterization of Strange Attractors. Phys. Rev. Lett. 1983, 50, 346–349.
[CrossRef]
56. Mandelbrot, B.B. How Long Is the Coast of Britain? Science 1967, 156, 636–638. [CrossRef] [PubMed]
57. Mandelbrot, B.B. The Fractal Geometry of Nature; Macmillan: New York, NY, USA, 1983.
58. Mandelbrot, B.B. Multifractals and 1/f Noise; Springer: New York, NY, USA, 1999.
59. Huikuri, H.V.; Mäkikallio, T.H.; Peng, C.K.; Goldberger, A.L.; Hintze, U.; Møller, M. Diamond Study Group.
Fractal Correlation Properties of RR Interval Dynamics and Mortality in Patients with Depressed Left
Ventricular Function after an Acute Myocardial Infarction. Circulation 2000, 101, 47–53. [CrossRef] [PubMed]
60. Takens, F. Detecting Strange Attractors in Turbulence. In Dynamical Systems and Turbulence; Rand, D.,
Young, L.S., Eds.; Springer: Berlin, Germany, 1981; pp. 366–381.
61. Theiler, J. Estimating Fractal Dimension. J. Opt. Soc. Am. A 1990, 7, 1055–1073. [CrossRef]
62. Kekre, H.B.; Sarode, T.; Halarnkar, P.N. A Study of Period Doubling in Logistic Map for Shift Parameter.
Int. J. Eng. Trends Technol. 2014, 13, 281–286. [CrossRef]
63. Clarke, K.C. Computation of the Fractal Dimension of Topographic Surfaces Using the Triangular Prism
Surface Area Method. Comput. Geosci. 1986, 12, 713–722. [CrossRef]
64. Sander, E.; Yorke, J.A. The Many Facets of Chaos. Int. J. Bifurc. Chaos 2015, 25, 1530011. [CrossRef]
65. Sprott, J.C.; Xiong, A. Classifying and Quantifying Basins of Attraction. Chaos 2015, 25, 083101. [CrossRef]
[PubMed]
66. Brown, T.A. Measuring Chaos Using the Lyapunov Exponent. In Chaos Theory in the Social Sciences; Kiel, L.D.,
Elliott, E., Eds.; University of Michigan Press: Ann Arbor, MI, USA, 1996; pp. 53–66.
67. Wolf, A.; Swift, J.B.; Swinney, H.L.; Vastano, J.A. Determining Lyapunov Exponents from a Time Series.
Phys. Nonlinear Phenom. 1985, 16, 285–317. [CrossRef]
68. Dingwell, J.B. Lyapunov Exponents. In Wiley Encyclopedia of Biomedical Engineering; Akay, M., Ed.; John Wiley
& Sons: Hoboken, NJ, USA, 2006.
69. Chan, K.S.; Tong, H. Chaos: A Statistical Perspective; Springer Science & Business Media: New York, NY,
USA, 2013.
70. Hunt, B.R.; Ott, E. Defining Chaos. Chaos 2015, 25, 097618. [CrossRef] [PubMed]
71. Kantz, H.; Radons, G.; Yang, H. The Problem of Spurious Lyapunov Exponents in Time Series Analysis and
Its Solution by Covariant Lyapunov Vectors. J. Phys. Math. Theor. 2013, 46, 254009. [CrossRef]

© 2016 by the author; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).