Leonard Schulman
California Institute of Technology, CA·Computer Science
Does a drug, a school policy or a regulation actually cause the effect we see, or is the link only a correlation? Randomized controlled trials are the gold standard for answering such questions, but ethics or cost often rule them out.
- Fields
Computer Science
Mathematics & Statistics
- Topics
causal inference
causal networks
causal discovery
- Grants
- 1 known grant, through Jun 2027
What they work on
- 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.
- Develop efficient causal inference methods for systems affected by several global confounders, hidden variables that each influence all observed quantities, rather than just one.
- Extend the causal network framework to the analysis of time series data, which is currently handled with a separate methodology.
- Prove strong information inequalities, limits on how information and correlation pass through a network, for causal networks with latent confounding and selection bias, generalizing strong data processing inequalities known for chains of noisy channels.
- Design causal discovery algorithms that infer network structure from statistical data efficiently and robustly to noise when a cardinality-bounded global confounder is present.
More about the research
Causal networks offer another route. These diagrams spell out which parts of a system cannot directly affect others. With the right assumptions, they can let researchers estimate causal effects from purely observational data. Current methods, however, work only under narrow assumptions and often do not scale to large networks. This theoretical computer science project aims to widen the range of situations where causal inference is possible. It will develop new algorithms, sample complexity bounds (how much data is needed) and limits on how strong correlations can be in large, sparse networks.
Why it matters
In public health, medicine, economics, education and regulation, randomized experiments are often unethical or impractical. More reliable and scalable methods for drawing causal conclusions from observational data could help researchers and policymakers in these fields judge what truly causes an outcome. The methods would also make clearer how far conclusions can be trusted when hidden factors or biased data collection are present.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Faramarz Fekri
Georgia Institute of Technology, GA·Computer Science
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.
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.
Nengkun Yu
Stony Brook University, NY·Computer Science
Determine the fundamental limits on how much information can be learned about a quantum system using Pauli measurements, a practical, widely used type of quantum measurement.
The school
California Institute of Technology: 18 labs on PI Pages, $445M a year on research, 200 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.