Reproduce “Koopman Generative Operators for Efficient Probabilistic Time-Series Forecasting” 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
Raz Marshanski, Liran Nochumsohn, Mayank Jauhari, Boris N. Oreshkin, Omri Azencot
KGO forecasts probabilistic time-series futures by encoding observed history, extending latent states with Koopman dynamics, and transforming noise into samples through a single closed-form transport. A deterministic forecast is blended with samples using variable- and horizon-conditioned weights. Across 17 ProbTS settings, KGO has lowest CRPS in 13 and lowest NMAE in 16; this result is limited to evaluated benchmarks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.