<ul><li>Stochastic Partial Differential Equations (SPDEs) driven by random noise are important for modeling rough spatio-temporal dynamics in various physical processes.</li><li>A new benchmark called SPDEBench has been introduced to efficiently model SPDEs of physical significance using machine learning methods on 1D or 2D tori.</li><li>The benchmark includes datasets for singular SPDEs based on the renormalization process and novel ML models that outperform traditional numerical solvers and existing ML-based models.</li><li>SPDEBench highlights the importance of considering computational errors introduced by noise sampling and renormalization when applying ML models to SPDE data, to avoid significant errors and misleading conclusions.</li></ul>

SPDEBench: An Extensive Benchmark for Learning Regular and Singular Stochastic PDEs

Discover more