My research in nonsmooth optimization and variational analysis connects the geometry of optimization models with the well-posedness and stability of their solutions. I develop concrete conditions, sensitivity bounds, and algorithms for regularized optimization, with applications in statistics and signal recovery.
Another major direction is maximum entropy on the mean (MEM), which turns probabilistic prior information into convex models for inverse problems. My work develops this framework’s foundations and computational methods, including stability and statistical guarantees when priors are learned from data.
- Nonsmooth optimization
- Variational analysis
- Inverse problems