Reproduce “Constraint- and Score-Based Nonlinear Granger Causality Discovery with Kernels” 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

Fiona Murphy, Alessio Benavoli

Researchers at Trinity College Dublin developed GP_SIC, a Gaussian Process-based method, and an improved Kernel Principal Component Regression (KPCR) for nonlinear Granger Causality, demonstrating superior performance in identifying both lagged and instantaneous causal links in time series data. GP_SIC consistently outperformed existing methods and was computationally efficient for contemporaneous causal discovery.

Key insights

  • Kernel Granger Causality and large-scale Nonlinear Granger Causality can be theoretically unified as specific instances of Kernel Principal Component Regression, providing a consolidated understanding of these methods.
  • Integrating the Smooth Information Criterion with Gaussian Processes' marginal likelihood significantly improves model selection, promoting sparsity and robustness in nonlinear causal discovery.
  • Granger Causality principles can be extended to infer instantaneous causal relationships using a GP_SIC-based algorithm, overcoming a traditional limitation while offering computational efficiency.

Summary by alphaXiv.

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