Score matching enables causal discovery of nonlinear additive noise models
About
This paper demonstrates how to recover causal graphs from the score of the data distribution in non-linear additive (Gaussian) noise models. Using score matching algorithms as a building block, we show how to design a new generation of scalable causal discovery methods. To showcase our approach, we also propose a new efficient method for approximating the score's Jacobian, enabling to recover the causal graph. Empirically, we find that the new algorithm, called SCORE, is competitive with state-of-the-art causal discovery methods while being significantly faster.
Paul Rolland, Volkan Cevher, Matth\"aus Kleindessner, Chris Russel, Bernhard Sch\"olkopf, Dominik Janzing, Francesco Locatello• 2022
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Causal Discovery | Sachs real-world data protein signaling network | SHD12 | 26 | |
| Causal Discovery | SF5 (n=30, h=5) | FDR55 | 18 | |
| Causal Discovery | ER5 (n=30, h=5) | FDR0.66 | 18 | |
| Causal Discovery | Synthetic ER3 n=50, h=3 (test) | FDR69 | 17 | |
| Causal Discovery | Synthetic SF3 n=50, h=3 (test) | FDR64 | 17 | |
| Causal Discovery | SF5 synthetic n = 50, h = 5 | FDR65 | 15 | |
| Causal Discovery | ER5 n = 50, h = 5 synthetic | FDR71 | 15 | |
| Causal Structure Learning | Scale-free (SF) datasets (n=100, h=5) synthetic (test) | FDR91 | 15 | |
| Causal Discovery | ER3 (n=100, h=3) Synthetic (test) | FDR91 | 15 | |
| Causal Structure Learning | Erdős–Rényi (ER) (n=100, h=5) synthetic (test) | FDR0.92 | 15 |
Showing 10 of 15 rows