Research

My research is highly interdisciplinary and intersects network science, signal processing, information theory, statistical machine learning, and others. In term of applications, I am currently interested in natural language processing (NLP), biological processes like population genetics modeling and phylodynamics, social processes like the spread of information or behavior.

I am interested in the properties of networks of interacting agents/components and developing analytical and computational tools to support the study of complex networks. Here are some currently funded as well as unfunded projects that my students and I are working on right now:

Project 1

Encoding and Decoding of Graph Structures and using Description Length for Model Selection and Anomaly Detection (Funded by NSF with Prof. Anders Høst-Madsen)

Encoding/decoding and compression of sequences are well-studied in information theory. However, many new data types are inherent graphs rather than sequences (e.g., social networks, genomic networks, etc.). Therefore, new methods need to be developed for graphs. What makes the problem challenging is that unlike sequences, graphs are, in general, unordered structures; that is, they lack a definite beginning and ending, which means that we can no longer use traditional, sequential based encoding or compression methods.

Coding can also be used to analyze the statistics of the data source. We explore methods of using graph encoding and description length analysis for graphical model selection, such as graphical lasso.

Project 2

Modeling Evolution of News using Dynamic Knowledge Graph Process (Funded by DARPA)

More news articles are written each day than a person can read. Natural language processing (NLP) algorithms teach computers to analyze a large amount of natural language data. One approach is to extract into object-relationship-object triplets from text. These triplets can be connected in the form of a directed multi-graph known as a knowledge graph.

Therefore, we can study a stream of news articles covering a particular event as a sequence of knowledge graphs. By analyzing the graph sequence, we can model the dynamics of news articles and detect fake news or anomalous information injection.

Project 3

Paremeter Estimation and Model Selection of Interacting Particle Systems

Interacting particle systems are continuous-time Markov processes that describe the stochastic evolution of node states on a graph structure. They are the continuous-time analog of stochastic cellular automata systems. Interacting particle systems have also been studied as dynamic Bayesian networks and dynamic Markov networks.

Complex system involving interacting particles/agents/states occurs separately in many different areas (e.g., quantum systems). The commonality is these systems are challenging to analyze. My current interest is in a feasible approach to parameter estimation and model selection of such a system.