Machine Learning Based Surrogate

Machine learning (ML)-based surrogates are computationally efficient approximations of complex physical simulations, aiming to drastically reduce computational costs while maintaining acceptable accuracy. Current research focuses on developing and evaluating these surrogates across diverse applications, including fluid dynamics, wave propagation, and complex system modeling, employing neural networks, Gaussian processes, and manifold learning techniques. This approach significantly accelerates design exploration, optimization, and uncertainty quantification in various scientific and engineering fields, enabling faster and more cost-effective analysis of complex systems. The effectiveness of these surrogates is being rigorously assessed through benchmarks and comparisons against traditional methods, with a particular emphasis on understanding the trade-off between data requirements and prediction accuracy.