About me

• I'm a second year PhD student in Computer Science Department of Stony Brook University. I'm advised by professo Jie Gao. I'm currently working on a project focused on modeling how information spreads between agents moving in a grid via different movement policies.
  
  Previously, I worked with Leman Akoglu, now at CMU.
  
  Check out my projects here.

Research Interests

• • Social and Information Networks
  • Large-Scale Graph Mining
  • Information/Disease Outbreak Analysis
  • Graph Theory

Education

• Stony Brook University
  PhD in Computer Science

  2015 - Present

• Sharif University of Technology
  B.Sc. in Computer Engineering

  2010 - 2015

Honors and Awards

• Renaissance Technology Fellowship

  2015 - 2017

  Awarded to select PhD students at Computer Science department of Stony Brook University.

• Silver Medalist in the 18^th Iranian National Olympiad

  2009

Peer Reviewed Publications

For an up-to-date list of my publications, please visit my Google Scholar page.

Conference and Journal Papers

• Ties That Bind: Characterizing Classes by Attributes and Social Ties

  Aria Rezaei, Bryan Perozzi, Leman Akoglu
  WWW Companion 2017, Perth, Australia
  [PDF] [Slides] [Code] [Data]

• Near linear-time community detection in networks with hardly detectable community structure

  Aria Rezaei, Saeed Mahloujifar, Mahdieh Soleymani
  ASONAM 2015, Paris, France
  [PDF] [Slides] [Code] [Data]

Workshop and Poster Papers

Nothing here yet.

Projects

• Characterizing Class Differences in Attributed Graphs

  Overview

  In social and information networks, attributes tend to strengthen the ties inside closely knit communities and weaken the ties between them and the world outside. Here, we propose a method to find such attributes in a setting where several classes of subgraphs are present and we find attributes that separate these classes from eachother.

• Controlled Label Propagation
  Preventing Ovevr-propagation through Gradual Expansion of Communities

  Overview

  Label Propagation Algorithm (LPA) is a lightweight, fast and secure method to retrieve community. However, due to over-propagation of labels during the early stages of the algorithm, some communities grow faster and bigger than they should and they flood the whole network. Here, we propose a simple and efficient heuristic to prevent the flooding. Our method can be used on top of most of the previously found heuristics.

Attributed Graphs

• Congress Dataset

  Attributed Undirected Temporal Node Meta Attribute Meta

  A group of congressmen who introduce a bill to the congress are called that bill's sponsors. Each bill has a group of policy terms. Here, there are 8 periods of congress which are turned into graphs where nodes are congressmen, edges are co-sponsorship relations and attributes are the policy terms a congressmen has introduced bills into.

• Amazon Co-purchaseship

  Attributed Undirected Node Meta Attribute Meta

  In this dataset, nodes are products and edges are between products that are frequently bought together. Here, attributes are properties about the genre, content or creators of the products. The products are also in either of the 4 categories of Books, Videos, Music or DVD.

• DBLP Co-authorship

  Attributed Undirected Node Meta Attribute Meta

  In this dataset, nodes are computer scientists on DBLP record and edges show co-authorship relation between to authors. Attributes are the venues of publication in Computer Science, famous conferences, journals and online repositories of scientific articles.