Reproduce “SCDM: Scalable Causal Discovery in Nonlinear Temporal Systems with Meta-Learning” 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
Jingbo Wang, Kegeng Tang, Zihao Wang, Shaogang Ren
SCDM discovers directed dependencies in nonlinear time series by sharing a predictor and source-to-target strength matrix across targets, training each update on a sampled subset. Episodes adapt copies before shared parameters update. On a 1,000-variable synthetic fMRI benchmark, at 800 seconds it reaches AUROC 0.645 and AUPRC enrichment 4.01×, exceeding CUTS+. This is controlled simulation evidence, not a general causal guarantee.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.