| In the Stochastic Analysis and Nonlinear Dynamics (SAND) lab our goal is to understand, predict, and/or optimize complex engineering and environmental systems where uncertainty or stochasticity is equally important with the dynamics. We specialize on the development of analytical, computational and data-driven methods for modeling high-dimensional nonlinear systems characterized by nonlinear energy transfers between dynamical components, broad energy spectra with complex statistics, and persistent or intermittent instabilities. We are particularly interested on the development of inexpensive predictive capacity for such systems as well as the development of design criteria for engineering applications. |
News
- Paper on active learning of neural operators has been published to Nature Computational Science
- NeurIPS paper recognized as Best Paper for ML innovation
- Paper on new ML training criteria for improving models for predicting extreme events has been accepted to Physica D
- Paper on data-driven prediction of vortex induced vibration is accepted to Ocean Engineering Journal
- Paper on optimal acquisition functions for active learning appears to the Philosophical Transactions of the Royal Society A
Featured Papers
 | Optimal acquisition functions for active learning of pdf tails T. Sapsis, A. Blanchard, Optimal criteria and their asymptotic form for data selection in data-driven reduced-order modeling with Gaussian process regression, Philosophical Transactions of the Royal Society A [pdf] |
 | Active learning with neural operators to quantify extreme events E. Pickering et al., Discovering and forecasting extreme events via active learning in neural operators, Nature Computational Science [pdf] |
 | A review of methods for modeling of extremes in fluids T. Sapsis, Statistics of extreme events in fluid flows and waves, Annual Review of Fluid Mechanics [pdf] |
 | Physically consistent NN closures for UQ of turbulent flows A. Charalampopoulos, T. Sapsis, Uncertainty quantification of turbulent systems via physically consistent and data-informed reduced-order models, Physics of Fluids [pdf] |