Title:
Network cascades: observations, models and algorithms
Speaker:
Jure Leskovec, CMU

Abstract: 
Network cascades are created by the diffusion process where
an action or idea like a virus spreads between the nodes of the
network. In this talk I will present our empirical observations,
models and algorithms on three real-world case studies of diffusion
and cascading behavior in networks.

First, I will consider cascades in a large viral marketing network of
16 million recommendatons, where people recommend products to each
other, and we studied the success and spread of recommendations over
the network. Next, we will examine patterns and models of the
propagation of information between a large population of blogs over a
long period of time. Last, we will focus on outbreak detection, where
given a water distribution network, the question is where to place
sensors in order to quickly detect contaminants. Or similarly, which
blogs should one read to avoid missing important stories? We frame the
problem as selecting nodes (sensor locations, blogs) in a network, in
order to detect the spreading of a virus or information as quickly as
possible.  We exploit ''submodularity'' to develop an efficient
algorithm that scales to large problems, achieving near optimal
placements, while being 700 times faster than a simple greedy
algorithm.


Bio: 
Jure Leskovec is a PhD candidate in Machine Learning Department,
School of Computer Science, at Carnegie Mellon University. He received
best paper awards at KDD '05 and '07 and won KDD Cup in 2003. Jure is
a Microsoft Research Fellow and also holds 3 patents. His research
interests include applied machine learning and large-scale data
mining, focusing on large real-world networks, their evolution, and
spread of information, influence and viruses over them.