(* indicates equal contributions)
False discovery control in non-standard settings
Modern scientific applications
consider vast and complex model spaces. As such, rigorous control of false discoveries is more
important than ever. Classical approaches, which consider simple decision rules, do not provide a framework for dealing with complex structures, such as phylogenetic trees in evolutionary biology, partitions over continuous data, and directed acyclic graphs in causal inference. Can we provide a unified framework for false discovery control in complex data structures?
- "Consensus Tree Estimation with False Discovery Rate Control via Partially Ordered Sets"
M. Valdez, A. Willis, A. Taeb
2026.
- "Quantifying Uncertainty and Stability Among Highly Correlated Predictors: a Subspace Perspective"
X. Zhang, J. Bien, A. Taeb [software]
2026.
- "Model Selection over Partially Ordered Sets"
A. Taeb , P. Bühlmann, V. Chandrasekaran [software][poster] [video].
Proceedings of National Academy of Sciences, 2024.
- "False Discovery and Its Control in Low-rank Estimation"
A. Taeb , P. Shah, V. Chandrasekaran, [software] [video].
Journal of the Royal Statistical Society (Series B), 2020.
Causal inference under uncertain graphical structure
Understanding causal relationships is a central goal of science. Causal graphs, such as directed acyclic graphs, are useful for representing these relationships. In practice, however, causal graphs are rarely known with certainty: they may be specified using incomplete background knowledge, learned from data, or only partially identified. Moreover, learning causal graphs involves challenging combinatorial optimization problems, where the lack of optimality guarantees can undermine the reliability of the resulting structures. How can we learn causal graphs with both provable optimality and statistical guarantees? How can we rigorously quantify uncertainty over causal structures? And how can we draw robust causal conclusions when the underlying graph is uncertain?
- "Interpretable confidence regions for causal discovery"
J. Kang, A. Mazumder, C. Lembo, A. Taeb
2026.
- "Model-oriented Graph Distances via Partially Ordered Sets"
A. Taeb*, F.R. Guo*, L. Henckel*
2026.
- "Convex Mixed-Integer Programming for Causal Additive Models with Optimization and Statistical Guarantees
"
X. Zhang, N. Keret, A. Shojaie, A. Taeb [software]
2026.
- "Characterization and Greedy Learning of Gaussian Structural Causal Models under Unknown Interventions
"
J. Gamella, A. Taeb , C. Heinze-Deml, P. Bühlmann, [software]
Journal of Machine Learning Research, 2026.
- "An Asymptotically Optimal Coordinate Descent Algorithm For Learning Bayesian Networks from
Gaussian Models
"
T. Xu, S. Küçükyavuz, A. Shojaie, A. Taeb [software]
Journal of Machine Learning Research, 2025.
- "Integer Programming for Learning Directed Acyclic Graphs from Non-identifiable Gaussian Models
"
T. Xu*, A. Taeb*, S. Küçükyavuz, A. Shojaie [software]
Biometrika, 2025.
- "A Fast Non-parametric Approach for Local Causal Structure Learning"
M. Azadkia, A. Taeb , P. Bühlmann
2022.
Learning from fragmented and heterogeneous data
Modern scientific applications routinely collect data from multiple sources or views. This heterogeneity creates
both challenges and opportunities for reliable scientific discovery. Different sources may contain
shared and source-specific latent factors, measure only partially overlapping sets of variables, exhibit
substantial distribution shifts, or differ most strongly in rare and extreme regimes. How can we
disentangle shared and source-specific latent structures? How can we reconstruct global relationships
from fragmented observations? How can we uncover invariant relationships across environments and
dependence structures in the extremes?
- "Convex Mixed-Integer Programming for Invariant Predictions from Multi-environment Data"
X. Zhang, R. Mazumder, A. Taeb ,
2026.
- "Integrating multi-source local information for global undirected graph reconstruction"
N. Keret*, A. Taeb*, and A. Shojaie
2026.
- "A Spectral Method for Multi-view Subspace Learning Using the Product of Projections
"
R. Sergazinov, A. Taeb , I. Gaynanova
Biometrika, 2025.
- "Causality-oriented Robustness: Exploiting General Noise Interventions"
X. Shen, P. Bühlmann, A. Taeb [software][poster]
Journal of American Statistical Association, 2025.
- "Extremal Graphical Modeling with Latent Variables via Convex Optimization"
S. Engelke, A. Taeb [software]
Journal of Machine Learning Research, 2025.
- "Learning and scoring Gaussian latent causal models with unknown additive interventions"
A. Taeb , J. Gamella, C. Heinze-Deml, P. Bühlmann, [software] [video]
Journal of Machine Learning Research, 2024.
- "Provable Concept Learning for Interpretable Predictions using
Variational Inference"
A. Taeb , N. Ruggeri, C. Schnuck, F. Yang, [software][video][poster]
ICML Workshop on AI4Science,2022.
- "Learning Exponential Family Graphical Models with Latent Variables using Regularized Conditional Likelihood"
A. Taeb , P. Shah, V. Chandrasekaran
2020.
- "Interpreting Latent Variables in Factor Models via Convex Optimization"
A. Taeb , V. Chandrasekaran
Mathematical Programming, 2018.
- "A Statistical Graphical Model of the California Reservoir Network"
A. Taeb , J.T. Reager, M. Turmon, V. Chandrasekaran [software] [press release] [broad audience]
Water Resources Research, 2017.
-
"Visual Stylometry using background selection and wavelet-HMT-based Fisher Information distances"
H. Qi, A. Taeb , and S. Hughes
EURASIP Signal Processing, 2012.