|
Networks: An Introduction
Mark Newman
Oxford University Press
784 pp. (2010)
ISBN: 978-0-19-920665-0
|
|
Reviewed by Luis H. Favela
Department of Philosophy
Center for Cognition, Action, and Perception, Department of Psychology
University of Cincinnati
Email: favelalh (at) mail.uc.edu
|
Network theory arguably has its origins in Euler’s (1741) graph theory, which was first
developed in the mid-1700s to solve the
Königsberg bridge problem.
Since then, the basic units of graph theory—vertices and edges—have been
utilized by a number of scientific disciplines to describe and analyze a wide
variety of phenomena. Mark Newman begins his clear and comprehensive
introduction to networks with a sampling of various kinds that have been
studied: information networks such as the World Wide Web, biological networks
such as neural connections, and social networks such as friendships among
members of a club. As Newman highlights, network theory and methods are utilized not only within but also across scientific disciplines. The
interdisciplinary nature of the scientific study of networks is a double-edged
sword. On the one hand, it fosters the cross-pollination of ideas and methods
across fields—for example, applying methods used to study computer networks to
the investigation of brain networks (cf., Sporns, 2011). On the other hand,
such interdisciplinarity can result in confusion and misapplication of theory
and methods. Accordingly, Newman’s goal in “Networks: An Introduction” is to
synthesize the current state of network theory into a “consistent language and
notation” (2010, p. x, Preface). This is no easy task, but Newman gives an admirable
attempt that is successful in many ways.
Newman begins with a brief introduction to basic network terminology, examples of
networks, and properties of networks (Chapter 1). The rest of the book is
divided into five parts. Part I (Chapters 2 to 5) consists of chapters
dedicated to the definition and description of four classes of real-world
networks: technological, social, information, and biological. Beyond explaining
in great detail the aforementioned classes of networks, this portion of the
book makes the case that network science is grounded in observational data. Thus,
although modeling plays a substantial role in network science, the end goal is
always to account for the behavior of real-world systems. Part II (Chapters 6
to 8) presents the quantitative foundations of network science. Whereas Part I
attempts to motivate the idea that countless phenomena can be understood as
networks (e.g., natural gas pipelines, citations between academic papers, and
collaborations of actors—i.e.,
“Six Degrees of Kevin Bacon”
[p. 54]), Part II begins to get into the nitty-gritty details of the conceptual and
mathematical representations of networks. For readers interested in applying
network theory, Part II provides a framework for researchers interested in
putting network theory into practice. Newman begins with a clear explanation of
how matrices and graphs lay the foundation for representations of networks. From
there, he builds up to more complex types of network representations as
hypergraphs and bipartite networks. While explaining the application of network
theory, Newman defines foundational terms used throughout the network
literature (e.g., degree, paths, components). Not only does Newman often pair
terminology with multiple explanations or examples, he often provides visual representations
of the concept in the forms of matrices and graphs as well as equations.
Part III (Chapters 9 to 11) discusses computational techniques and algorithms for
analyzing networks. Chapter 9 discusses foundational issues related to
analyzing networks computationally. It is here that one of the most important
challenges of applying network theory arises, stemming from the
interdisciplinary application of network science mentioned above. Newman claims
that, “[m]uch time can be wasted when people fail to understand how a program
works or misunderstand the kinds of answers the program can give them” (2010,
p. 277). Just because the tools of network science can be readily applied to a
phenomenon does not mean that just any kind of program or analysis will be
appropriate. Not only can failures to correctly apply algorithms result in
misinformation, it can lead to a drastic inefficiency in data analysis, as in
the case of computing eigenvector centrality (pp. 345-354). Eigenvector
centrality of vertex i in a network
is defined as the ith component of
the adjacency matrix’s leading eigenvector (p. 345). There are numerous ways to
calculate the eigenvector. However, as Newman
points out, such calculations can include a lot of unnecessary data and be
immensely time-consuming. Treating eigenvector centrality with the power method is
simpler, faster, and just as reliable as the standard linear methods (pp.
346-347). If a researcher is not aware of the mathematics behind the
computational methods, then it is possible that she could waste time applying a
relatively inefficient method. It is for that reason that Newman spends a
substantial portion of the book going step-by-step to help the reader
understand the algorithms at work in the computational software. Such
understanding can help a researcher apply the appropriate tools and develop
their own programs.
With the conceptual and methodological foundation put in place in Parts I to III,
Part IV (Chapters 12 to 15) reviews specific models of network structure and
patterns of connectivity, such as various types of random graphs (Chapters 12
to 13) and generative networks (Chapter 14), and more recent models such as
small-world networks (Chapter 15). The final part of the book, Part V (Chapters
16 to 19), reviews various kinds of behaviors exhibited by networks. Whereas
the previous chapters discuss network structures, the final chapters discuss
the processes exhibited by those systems. Processes discussed include
percolation and resilience (Chapter 16), epidemics (Chapter 17), dynamical
(Chapter 18), and searching (Chapter 19). The material covered in Parts I to IV
demonstrate that network science has a strong grasp of network structure.
However, as Newman draws attention to, the processes of these systems are far
from understood—such as nonlinear dynamics in brain networks (cf. Sporns,
2011).
“Networks: An Introduction” has a few, minor shortcomings. First, as an introductory text,
it would be useful to have a glossary of network terminology and classes, as
well as a list of equations and models for the different kinds of networks.
Second, although it is useful to have problems at the end of each chapter, it
would be more convenient to have solutions within the text instead of a
separate solutions manual. Third, the abrupt ending of the book after a chapter on network
search processes leaves room for a concluding chapter. A concluding chapter
would be useful to review how the author accomplished the goal set out at the
start of the book, namely, to synthesize the current state of network science
into a “consistent language and notation” (Newman, 2010, p. x, Preface). Finally,
although I can understand why Newman would not want to bias the reader with any
one particular kind of software, some readers, especially the more novice to
the field of network science, might want more concrete suggestions and sources
for software tools (e.g., “Using your favorite numerical software...” [p.
392]). These shortcomings are very minor and do not detract from Newman’s
achievement: an introduction to the burgeoning field of network science that is
clear and easy to read and succeeds in presenting network theory and methods in a
consistent language and notation. This book is appropriate for advanced
undergraduate and graduate students from various disciplines interested in both
the theory and application of network science. Newman has set the standard on
which other introductions to network science will be measured.
References
Euler, L. (1741). Solutio problematis ad geometriam situs pertinentis
[The solution of a problem relating to the geometry of position)].
Commentarii academiae scientiarum Petropolitanae, 8, 128-140.
Sporns, O. (2011). Networks of the brain. Cambridge, MA: MIT Press.