Reproduce “AI Feynman 2.0: Pareto-optimal symbolic regression exploiting graph modularity” 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
Silviu-Marian Udrescu, Andrew Tan, Jiahai Feng, Orisvaldo Neto, Tailin Wu, Max Tegmark
AI Feynman 2.0 presents an improved method for symbolic regression that exhibits orders of magnitude greater robustness to noise and successfully discovers more complex mathematical formulas. The approach systematically exploits generalized graph modularity identified through neural network gradients and applies a Pareto-optimal search using information-theoretic metrics, also extending symbolic regression to probability distributions from samples.
Key insights
- Generalized graph modularity, including compositionality and generalized symmetry, can be systematically detected from neural network gradient properties, enabling recursive decomposition of complex problems.
- A Pareto-optimal search, guided by information-theoretic metrics like Mean Error-Description-Length (MEDL) for accuracy and description length (Ld) for complexity, provides robustness to outliers and inherently prevents overfitting.
- Statistical hypothesis testing can efficiently prune the vast search space of symbolic expressions, significantly accelerating the discovery process for base cases and modular components.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.