J. Shanghai Jiao Tong Univ. (Sci.), 2017, 22(1): 87-91
DOI: 10.1007/s12204-017-1805-9
Continuous Multiplicative Attribute Graph Model
HUANG Jiaxuan (), JIN Xiaogang
∗
()
(Institute of Artifical Intelligence, College of Computer Science and Technology, Zhejiang University,
Hangzhou 310027, China)
© Shanghai Jiao Tong University and Springer-Verlag Berlin Heidelberg 2017
Abstract: Network modeling is an important approach in many fields in analyzing complex systems. Recently
new series of methods have emerged, by using Kronecker product and similar tools to model real systems. One of
such approaches is the multiplicative attribute graph (MAG) model, which generates networks based on category
attributes of nodes. In this paper we try to extend this model into a continuous one, give an overview of its
properties, and discuss some special cases related to real-world networks, as well as the influence of attribute
distribution and affinity function respectively.
Key words: multiplicative attribute graph model, social network, continuous attribute
CLC number: TP 181 Document code: A
0 Introduction
Complex systems can be observed in multiple fields,
ranging from ecosystems and metabolic networks to so-
cial and economic systems, and networks provide a use-
ful abstraction of such systems
[1-5]
. In the past several
decades there have been significant advances in ana-
lyzing statistical properties and topological structures
of network-liked systems
[6-9]
, but there remain lots of
challenges in extracting scientific understanding from
the large quantities of data produced by the experi-
ments and from those real-world systems
[10]
.
Social networks are networks of social relationships
or interactions, in which the vertices are people or
sometimes groups of people, and the edges are social
relationships between vertices, such as friendship and
fellowship
[11]
. Social networks have become very popu-
lar in recent years, because of the increasing popularity
and spread of internet enabled devices such as personal
computers, mobile devices and other hardware inno-
vations. And a number of important problems have
then arisen in the context of structural analysis of so-
cial networks. One such line of research is to try to
understand and model the nature of those very large
online networks
[12-13]
. Since a larger amount of data
are available for the case of online social networks, the
verification of such structural properties is much more
Received date: 2016-07-14
Foundation item: the National Natural Science Foun-
dation of China (No. 61379074) and the Zhejiang
Provincial Natural Science Foundation of China
(No. LZ12F02003)
∗E-mail: xiaogangj@cise.zju.edu.cn
robust in terms of statistical significance. Using the
massive amounts of data, it has actually become possi-
ble to study classical characters in detail
[14]
.
Many traditional network models investigate edge
creation mechanisms based on network structure only,
but actually a set of attributes is associated with each
node as well in the real world. This is especially true in
social networks, where not only relationships between
people, but also their own characteristics and profile in-
formation is available. In this case, both characteristics
of nodes and the network structure should be consid-
ered equally.
In order to capture the interaction between the net-
work structure and node attributes, we attempt to ex-
tend the multiplicative attribute graph model, makes
it available to combine continuous attributes of vertices
into the generating process of networks. We consider a
model in which each node has a vector of continuous
attributes associated with it, and the linking probabil-
ity of each pair of nodes is related to their attribute
vectors.
1 Recent Works
To study the characteristics and structures of real
systems, different network models have been proposed.
Though no models so far can perfectly match the real
world, they still provided some interesting views of net-
works. Recently a new line of work has emerged, de-
veloping network models that are analytically tractable
since one can mathematically analyze structural prop-
erties of networks produced from the models, as well
as statistically meaning since they also have efficient