Code
Research code accompanying our papers, developed in the Uncertainty Quantification Group and released under the UQUH organization. Each repository reproduces the main experiments of its paper.
SS-PPCA
A parametric distribution over reduced subspaces for characterizing model error in high-dimensional simulations.
MATLAB R2023b or later · Statistics and Machine Learning Toolbox · Optimization Toolbox
SS-Bootstrap
The nonparametric counterpart: stochastic subspaces drawn from the empirical data distribution via the bootstrap, dropping the Gaussian assumption.
MATLAB R2023b or later · Statistics and Machine Learning Toolbox · Optimization Toolbox
SO-BO-scale
Bayesian optimization under uncertainty for tuning a scale parameter when every evaluation is noisy. Ships with GP and Monte Carlo baselines, state-of-the-art noisy-BO baselines, and a robustness study across seeds.
MATLAB R2023b or later · Python (numpy, scipy, matplotlib)