Surrogate modeling involves creating simplified models, often based on data-driven or statistical techniques, to approximate the behavior of complex computational or physical systems. These models, such as polynomial regressions, Gaussian processes, or neural networks, significantly reduce computational costs while preserving accuracy for tasks like optimization, sensitivity analysis, and uncertainty quantification.
Features in QUEENS
- Polynomial Regression
- Gaussian Processes Regression (Kriging)
- Neural Networks
- Bayesian Neural Networks
- Polynomial Chaos Expansions