<ul><li>A new research paper titled 'Autoregressive regularized score-based diffusion models for multi-scenarios fluid flow prediction' has been published on arXiv (ID: 2505.24145v1).</li><li>The paper introduces a conditional score-based diffusion model for predicting fluid flow in multiple scenarios, combining scientific machine learning and generative modeling techniques.</li><li>The model includes an energy constraint based on turbulent flow statistical properties to enhance prediction quality with minimal training and cost-effective sampling.</li><li>Extensive experiments on complex fluid dynamics datasets show that the proposed model consistently provides stable, robust, and accurate predictions across different flow regimes and configurations, maintaining key physical and statistical properties.</li></ul>

Autoregressive regularized score-based diffusion models for multi-scenarios fluid flow prediction

Discover more