age shapes Web
Technology Research News
As it turns out, networks like the Web
have something in common with humans -- how they are structured has a
lot to do with aging.
Two researchers from the Niels Bohr Institute in Denmark have found they
can more accurately model networks like the Web
by using a sort of memory that takes into account the potential for
waning popularity as a Web page gets old. They also found that this age
variable plays a part in the stability of real world networks.
This is a step forward from earlier models, which tend to represent popularity
simply as the number of links a node has accumulated.
The older a Web page is the more likely it is to go out of date and become
less popular, said Konstantin Klemm, a graduate student at the Institute.
"After an exciting Web page has been launched, many other pages link to
it. However, sooner or later the novelty and consequently the popularity
of this page decreases. Interest shifts to pages released more recently,
which are now the ones that rapidly increase the number of incoming links."
The age variable essentially introduces time into the network structure
model, said Klemm. This makes for more accurate simulations of real world
networks, which can potentially help in predicting how networks like the
Web and wireless telecommunications will grow, he said.
It also stabilizes the network, said Klemm. "This gives insight into which
growth mechanisms lead to stable structures, which can be very useful
in devising strategies for expanding telecommunications networks," he
The model gives each node a certain amount of staying power based on how
novel its contents are. "Each node in the network is assigned an extra
degree of freedom containing the ability to attract links -- [its] popularity,"
said Klemm. When a node is created it is linked to nodes that are popular
at the time. It then receives links from nodes created subsequently, he
said. "This continues until eventually the node under consideration loses
The model mimics the appearance of new pages on the Web, which have links
to popular pages instead of outdated, forgotten ones, Klemm said. "A Web
page released at the beginning of the year 2002 will most likely contain
many links to pages of the year 2001, [and] fewer references to 2000,"
he said. The researchers made the assumption that the more links point
to a given Web page, the less likely it will be forgotten, said Klemm.
When the researchers tested their model, they found it more closely maped
real-world networks than previous ones. Many networks, including the Web,
have several structural traits in common.
Scale-free networks have a few nodes with many links and many nodes with
only a few links.
In small-world, or six-degrees-of-separation networks, every node can
reach every other node through a relatively small number of hops, or jumps
from one node to another. This small-world trait emerges when nodes in
a network become clustered into groups, and is also true of networks of
people, who cluster according to dynamics like where they live, work,
or what interests they have.
Many network models show how either the small-world or the scale-free
structure arises. Both structures emerge from the researchers' simulations,
according to Klemm. "Our work connects the concepts of growing scale-free
networks and the small-world effect. In our model the crucial properties
of the network emerge from one simple continuous growth process."
Clustering develops in the model due to the age variable, he said. "This...
locality in time... causes clustering," he added.
The work is "an important step ahead in network modeling," said Reka Albert,
a research associate at the University of Minnesota. There are two major
directions in network modeling: small-world networks and scale-free networks,
she said. The Niels Bohr Institute model "unites the large clustering
coefficient of small-world network models and the... distribution of scale-free
The model is useful because it closely maps real world networks, many
of which have at least some amount of node memory, said Albert. "In citation
networks there's a clear tendency to refer to more recent papers. In collaboration
networks... new collaborations are restricted to the active people."
The Web is also affected by some degree of node memory, but the situation
is not as clear-cut, Albert said. This is because it not only contains
both pages with unchanged content that become outdated and forgotten,
but also well-maintained pages that receive new links independent of their
creation time, she said.
The next step in the research is to study dynamics like communication
between nodes and propagation of diseases in their network models, said
Klemm. The research should eventually point the way toward better search
strategies on the Web. The idea could be applied practically within five
years, he said.
The research was funded by The Danish National Research Council.
Timeline: < 5 years
TRN Categories: Networking; Internet
Story Type: News
Related Elements: Technical paper, "Highly Clustered Scale-Free
Networks," posted in the arXiv physics archive http://arXiv.org/abs/cond-mat/0107606.
Chemists create nano
English could snowball
Page age shapes Web
six degrees of separation
Spot of gold makes
Research News Roundup
Research Watch blog
View from the High Ground Q&A
How It Works
News | Blog
Buy an ad link