Research
differential networks
Estimating and testing how the conditional-dependence structure of a system changes between two conditions, in high dimensions.
- Working paper Assumption-lean inference for spectral differential network analysis of high-dimensional time series
- Presentation Statistical inference for high-dimensional spectral differential networks
- Paper Spectral differential network analysis for high-dimensional time series
- Software KLIEPInference.jl
- Paper Two-sample inference for high-dimensional Markov networks
predictive inference for AI/ML models
Distribution-free predictive confidence sets — conformal prediction and the jackknife+-after-bootstrap — for arbitrary learning algorithms.
- Presentation Invited discussant, International Seminar on Selective Inference
- Presentation Predictive inference is free with the jackknife+-after-bootstrap
- Paper Predictive inference is free with the jackknife+-after-bootstrap
fundamental limits of black-box inference
What can, and cannot, be certified about a black-box algorithm from data alone — for example, tests for algorithmic stability.
- Presentation Black-box tests for algorithmic stability
- Paper Black-box tests for algorithmic stability
multivariate time series
Inference for high-dimensional, dependent data, including spectral differential-network analysis of time series.
- Working paper Assumption-lean inference for spectral differential network analysis of high-dimensional time series
- Presentation Statistical inference for high-dimensional spectral differential networks
- Paper Spectral differential network analysis for high-dimensional time series
interdisciplinary collaborations
Methods developed with domain scientists, including accelerated breast-MRI reconstruction.
- Software Enhancement-constrained acceleration (ECA)
- Paper Enhancement-constrained acceleration: A robust reconstruction framework in breast DCE-MRI
- Presentation MRI reconstruction