Reproduce “SDECast: Probabilistic Weather Forecasting in Continuous Time with Neural SDEs” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Maria Marchenko, Martin Andrae, Fredrik Lindsten, Christian A. Naesseth
SDECast models weather as a continuous-time stochastic process, using drift for systematic change and diffusion for variation across forecast paths. It adapts simulation-free SDE Matching to anchor learned states to physical observations. On hourly ERA5, its 24-hour t850 RMSE is 1.44 K, below AR-1h at 1.69 K, but above ARCI-24-1h at 1.26 K and comparable to DYffusion at 1.43 K.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.