University of Southern California
ISI Site Signature

University of Southern California


Aram Galstyan
 
 
       
  Modeling Complex Networks  
  Networks are inherently complex dynamical systems, where both attributes of individuals (nodes) and topology of the network (links) can have inter-coupled dynamics. For instance, it is known that social networks tend to divide into groups, or communities, of like-minded individuals. One can ask whether individuals become likeminded because they are connected via the network, or whether they form network connections  
       
  To model the interplay between influence and selction, we have developed a computational models of co-evolving networks based on interacting Hidden Markov Processes. The idea behind this approach is that he network is shaped by the interaction of local dynamical processes unfolding on individual nodes, while those processes themselves are influenced by the changing network structure. This provides a feedback mechanism that is vital for capturing realistic behavior of complex real-world network.  
       
  Another model of co-evolving dynamics is based on agents involved in game theoretical interactions. Specifically, we consider network-augmented multi-agent systems where, at each time step, agents choose which neighbor to play with, and which strategy to choose. As agents play repeatedly with each other, they will adapt their behavior by reinforcing (penalizing) both links and strategies that provide good (bad) outcomes. Thus, the reinforcment affects both the strategy and network evolution.  
       
   
       

       
  Papers  
       
  Ardeshir Kianercy and Aram Galstyan, Dynamics of Softmax Q-Learning in Two-Player Two-Action Games, in submission, 2011.

 
  Greg Ver Steeg, Aram Galstyan, and Armen Allahverdyan Statistical Mechanics of Semi-Supervised Clustering in Sparse Graphs, submitted to JSTAT, accepted for a poster presentation at UAI-11, 2011.

 
  Yoon Sik Cho, Greg Ver Steeg and Aram Galstyan, Co-evolution of Selection and Influence in Social Networks, In Proc. of the Twenty-Fifth Conference on Artificial Intelligence (AAAI-11), 2011.

 
  Ardeshir Kianercy and Aram Galstyan, Replicator Dynamics of Co-Evolving Networks, In AAAI Fall Symposium on Complex Adaptive Systems, Nov. 2010.
 
       
  Project Staff  
       
  Aram Galstyan (PI)
Greg Ver Steeg
Ardeshir Kianercy
 
       
       
  This research is sponsored in part by the National Science Foundation under Award No 0916534.  
Background