Faramarz Fekri
Georgia Institute of Technology, GA·Computer Science
Spotting that two things happen together is easy. Knowing whether one causes the other is much harder, yet it is the knowledge needed to decide what to change in medicine, economics, engineering or public policy.
- Fields
Computer Science
Mathematics & Statistics
Engineering
- Topics
causal inference
causal discovery
causal graph learning
- Grants
- 1 known grant, through Sep 2028
What they work on
- Identify when cause-and-effect relationships can be calculated from observed data even when unmeasured confounders (hidden variables that affect both cause and outcome) are present, and develop robust ways to estimate those effects.
- Develop scalable, optimization-based differentiable methods that learn causal graphs, diagrams of which variables influence which, from data that are noisy or partially missing.
- Build a framework that uses large language models (LLMs) to extract knowledge from scientific literature and databases, then uses it to validate or revise causal models when observational data are sparse.
More about the research
Machine learning is good at finding patterns and correlations, but those patterns cannot say why events happen or what would follow from an intervention. Causal discovery tries to uncover the cause-and-effect structure behind data. Real datasets are often large, noisy, incomplete or simply too small, and that makes the job difficult.
Why it matters
Better tools for finding causes in messy, incomplete data could help doctors, economists, engineers and policymakers predict what an intervention will actually do, not just what tends to occur alongside it. More interpretable causal models could support targeted decisions in healthcare and public policy where data are limited or contain hidden factors.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Razieh Nabi
Emory University, GA·Mathematics & Statistics
Identify causal effects as functions of observed data and estimate them robustly when unmeasured confounders — hidden variables that influence both the cause and the outcome — are present.
Leonard Schulman
California Institute of Technology, CA·Computer Science
Relax the statistical separation assumption needed when a hidden global confounder takes a limited number of values, aiming instead for model identification measured in Wasserstein distance, a way of comparing probability distributions.
Kasim Candan
Arizona State University Campus Immersion, AZ·Computer Science
Develop PanAX methods that separate setting-specific from transferable parts of epidemic data and models, so knowledge from data-rich outbreaks can be reused for new outbreaks with limited data.
The school
Georgia Institute of Technology: 50 labs on PI Pages, $1.5B a year on research, 572 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.