Reproduce “Filling the G_ap_s: Multivariate Time Series Imputation by Graph Neural Networks” 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

Andrea Cini, Ivan Marisca, Cesare Alippi

GRIN imputes missing readings in multivariate time series by modeling sensors as graph-connected nodes. Bidirectional recurrent graph modules combine information across time with message passing between related sensors, and a neighbor-constrained decoder refines estimates. On PEMS-BAY point-missing data, GRIN reports MAE 0.67 versus 1.47 for BRITS; this benchmark result does not establish performance for arbitrary missingness.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.