According to Essential Science Indicators
fromThomson
Reuters, the paper "Emergence of scaling in
random networks" (Barabási AL, Albert R,
Science 286[5439]: 509-12, 15 October 1999) ranks at #5
among Highly Cited Papers in the field of Physics,
garnering 2,708 citations between January 1, 1998 and
August 31, 2008.
The paper's authors are Albert-László
Barabási and Réka Albert. Dr.
Barabási is the Distinguished University
Professor of Physics and the Director of the Center for
Complex Network Research at Northeastern University in
Boston, Massachusetts, and holds an appointment at the
Department of Medicine, Harvard Medical School. Dr.
Albert is an Associate Professor of both Physics and
Biology at the Pennsylvania State University, where she
is also a member of the Center for Infectious Disease
Dynamics.
This month, ScienceWatch.com
talks with both authors about their paper and its
influence.
What was your inspiration for this study
reported in your paper?
Our earlier study, published in the same year in Nature (Albert R,
Jeong H, Barabási AL, "Internet—diameter of the World-Wide
Web," Nature 401[6749]: 130-1, 9 September 1999), has shown that
the World Wide Web is a not a random network, but the number of links per
node, often called the degree distribution, follows a power law. This was
rather unexpected, as earlier network models predicted only Poisson
distributions in this context. Subsequently we found that not only the WWW,
but other networks, like the actor network or the citation network, follow
the same distribution. These different datasets together indicated that we
are dealing with a potentially universal behavior, which might have a
common explanation. This paper was intended to offer that missing
explanation.
How did you perform this study—what were
your methods?
Coauthor:
Réka Albert
This is a theoretical paper, and its essence is a model that we proposed to
explain the origin of the power laws seen in various complex networks. The
model, based on the idea that networks grow by the addition of new nodes,
and these new nodes have a tendency to link to more connected nodes
(preferential attachment), was first simulated numerically, then we
obtained an analytic solution, predicting the observed power law
distribution.
Would you sum up your findings for our
readers?
The novelty of the proposed model was that for the first time it considered
the network as a dynamical object, that evolves with the addition of nodes
and links to the system, in strong contrast with the static models that
dominated the literature before. As such it faithfully captured the
emergence of real networks, predicting their large-scale topology
relatively accurately.
How was this paper received by the
community?
At first the community was slow to pick it up, as the paper was not solving
an established problem in a well-defined field with a strong preexisting
community, but instead opened a new field. However, soon papers started to
emerge in two areas: First, they provided evidence that many other networks
are scale-free (which is the name of the network that our model generated),
expanding the applicability of our results. Second, several investigators
offered extensions of the model, showing that it can account for the
structure of many real networks. Only about two years later the paper
entered its "exponential" phase, an indication that the amount of work in
this area exploded.
Where have you taken your research since the
publication of the 1999 paper?
In 2000 we found that the scale-free structure proposed in this paper
applies to the cell as well, characterizing its metabolic network (Jeong H,
et al., "The large-scale organization of metabolic networks,"
Nature 407[6804]: 651-4, 5 October 2000) and protein interaction
network (Jeong H, et al., "Lethality and centrality in protein
networks," Nature 411[6833]: 41-2, 3 May 2001). We have also shown
that scale-free networks, which due to the power law distribution are
dominated by hubs, show a high degree of robustness to errors, but are
fragile to attacks (Albert R, Jeong H, Barabási AL, "Error and
attack tolerance of complex networks," Nature 406[6794]: 378-82,
27 July 2000).
"The novelty of the proposed model was
that for the first time it considered the network
as a dynamical object..."
Subsequently many other researchers have demonstrated a series of
fascinating properties of these networks, finding not only that a very
large number of real networks, from the Internet to social networks, are
scale-free, but also that this scale-free topology fundamentally changes
the system's behavior, affecting everything from spreading processes on
networks (like the spread of infectious diseases) to cascading failures.
What are your hopes for this field for the
future?
Since 1999 a rather active highly interdisciplinary field was born, which a
recent National Research Council report dubbed Network Science, bringing
physicists, computer scientists, biologists, social scientists, and
researchers from many other disciplines under the same umbrella. These days
there are several conferences each year devoted to the topic, and several
dozen books have appeared on the subject. Networks offer a mathematically
rigorous formalism that is critical for understanding complex systems,
explaining the reason why it has been so embraced by various communities.
The role of the statistical physics community has been particularly
important, and we consider network science an integral part of statistical
physics these days, as it is dominated by the tools of statistical
mechanics, and has also led to many fascinating and challenging problems
that have captured the physics community. We very much hope that this
enthusiasm will continue, eventually leading to a rigorous treatment of
networked systems, from cells to social systems.
Albert-László Barabási
Director of the Center for Complex Network Research
Department of Physics
Northeastern University
Boston, MA, USA
Réka Albert
Department of Physics
Department of Biology
The Pennsylvania State University
University Park, PA, USA
Albert-László
Barabási and Réka Albert's current most-cited
paper in
Essential Science
Indicators, with
2,708 cites: