Inverse UQ

Inverse uncertainty quantification (UQ) aims to determine the uncertainty in model inputs based on observed experimental data, particularly focusing on efficiently handling complex, high-dimensional data like time-series. Current research emphasizes the use of dimensionality reduction techniques, such as functional principal component analysis, coupled with surrogate models like deep neural networks to improve computational efficiency in Bayesian inference frameworks. This approach is proving valuable in various applications, enabling more accurate and efficient parameter estimation and uncertainty characterization in complex systems, such as those modeled by physical simulations.