Research map / Probabilistic and scientific modeling / Causal discovery

Causal-learn: Causal Discovery in Python

NeurIPS 2024 · Probabilistic and scientific modeling › Causal discovery · approach: Constraint-based structure learning with latents · 26 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Describes causal-learn, an open-source pure-Python library offering a broad collection of causal discovery algorithms, independence tests, score functions, and utilities.

Problem

Causal discovery tools are mostly in Java or R (TETRAD, pcalg, bnlearn), which complicates deployment for Python-focused machine learning users and lacks a unified Python package covering all major method categories.

Approach

A fully Python, modular library covering constraint-based, score-based, functional causal model-based, and latent-variable causal discovery methods, with independent modules for independence tests, score functions, graph operations, and evaluation metrics.

Key results

• Provides a pure-Python implementation of PC, FCI, GES, CD-NOD, DirectLiNGAM, VAR-LiNGAM, RCD, CAM-UV, GIN and other methods, installable via pip.
• Includes independence tests (Fisher-z, Chi-square, G-square, KCI), score functions (BIC, BDeu, generalized score), and evaluation metrics such as SHD.
• Offers a one-line API example: pc(data) followed by graph visualization.

Limitations

• Constraint-based methods often return only Markov equivalence classes with undirected edges.
• Granger causality is statistical rather than true causal inference under stated assumptions.

Experimental setup

Baselines: TETRAD pcalg bnlearn py-tetrad py-causal Causal Discovery Toolbox

Builds on: PC FCI GES LiNGAM GIN

Closest papers by meaning

• Integrating Background Knowledge for Scalable Causal Discovery (NeurIPS 2026 · similarity 0.83)
• A Meta-Learning Approach to Bayesian Causal Discovery (ICLR 2025 · similarity 0.83)
• THE ROBUSTNESS OF DIFFERENTIABLE CAUSAL DISCOVERY IN MISSPECIFIED SCENARIOS (ICLR 2025 · similarity 0.83)
• CausalTab: Pretraining Across Causal Environments for Tabular Causal Discovery (NeurIPS 2026 · similarity 0.83)
• Adaptive Online Experimental Design for Causal Discovery (ICML 2024 · similarity 0.83)
• On Causal Discovery in the Presence of Deterministic Relations (NeurIPS 2024 · similarity 0.83)