Reproduce “Gradient-Based Neural DAG 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
Sébastien Lachapelle, Philippe Brouillard, Tristan Deleu, Simon Lacoste-Julien
GraN-DAG learns nonlinear causal graphs by modeling each variable’s conditional distribution with a neural network and enforcing acyclicity through a smooth constraint. On synthetic 10-node ER4 data, it achieves mean SHD 8.3 and SID 21.8 across 10 datasets of 1,000 samples each, outperforming CAM on both measures. Across tasks CAM sometimes does better; recovery depends on identifiability assumptions and optimization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.