Stat Cafe - Trisha Dawn
Covariate-assisted graph matching
- Time: Tuesday 12/2/2025 from 11:10 AM to 12:25 PM
- Location: BLOC 448
Description
Data integration requires aligning multiple data sources often without unique identifiers. For network data, a crucial initial step in this process is performed through graph matching. The covariates associated with nodes or edges can be instrumental in achieving improved accuracy. However, most existing techniques ignore this information, limiting their performance against non-identifiable and erroneous matches. We propose two novel covariate-assisted seeded graph matching approaches, based on quadratic assignment problem and leveraging local neighborhood structure, respectively. Based on a conditional modeling framework using GLM, we provide theoretical guarantees for model estimation and exact recovery. We validate the performances through simulations and demonstrate improved alignment on matching academic genealogy and collaboration networks.
Our Speaker
Trisha Dawn is currently a doctoral candidate at the Department of Statistics at Texas A&M University, being advised by Dr. Yang Ni and Dr. Jesus Arroyo. She received her master’s degree in Statistics from Indian Statistical Institute. Trisha’s research is broadly focused on the intersection of developing novel statistical methods and applications for causal graph learning, graph matching and change-point detection.