Advances in Flow Modeling and Simulation

The field of flow modeling and simulation is experiencing significant developments, driven by innovative applications of machine learning and mathematical techniques. Researchers are exploring new methods to improve the accuracy and efficiency of flow simulations, including the use of neural surrogates, sparse identification, and stochastic modeling. These advances have the potential to impact various fields, such as aerospace, automotive, and renewable energy. Notably, papers on importance-weighted non-IID sampling and sparse Kalman identification have introduced novel frameworks for estimating expectations and identifying model structures. Another noteworthy paper proposes a neural surrogate model for designing gravitational wave detectors, demonstrating the potential for machine learning to accelerate scientific discovery. Additionally, research on stellarator design using generative AI and sparse-to-field reconstruction via stochastic neural dynamic mode decomposition showcases the growing intersection of machine learning and physics. Some particularly noteworthy papers include: Importance-Weighted Non-IID Sampling for Flow Matching Models, which introduces a novel framework for estimating expectations. Sparse Kalman Identification for Partially Observable Systems via Adaptive Bayesian Learning, which achieves accurate model structure selection with high efficiency. Neural surrogates for designing gravitational wave detectors, which rapidly predicts the quality and feasibility of candidate designs.

Sources

Importance-Weighted Non-IID Sampling for Flow Matching Models

Sparse Kalman Identification for Partially Observable Systems via Adaptive Bayesian Learning

Sparse Broad Learning System via Sequential Threshold Least-Squares for Nonlinear System Identification under Noise

Neural surrogates for designing gravitational wave detectors

Diffusion for Fusion: Designing Stellarators with Generative AI

Sparse-to-Field Reconstruction via Stochastic Neural Dynamic Mode Decomposition

Autoregressive Surrogate Modeling of the Solar Wind with Spherical Fourier Neural Operator

Going with the Speed of Sound: Pushing Neural Surrogates into Highly-turbulent Transonic Regimes

Built with on top of