Reproduce “Probabilistic Learning of Multivariate Time Series with Temporal Irregularity” 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
Yijun Li, Cheuk Hang Leung, Qi Wu
RFN forecasts irregular multivariate time series by combining a hidden-state model with a conditional continuous normalizing flow (CNF). The flow transforms a Gaussian base into a forecast distribution, while a masked likelihood excludes unobserved variables. On asynchronous MuJoCo data with 50% missingness, mean CRPS was 0.1970 ± 0.0052, versus 0.2695 ± 0.0054 for GRUODE across five experiments. This result applies to that test setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.