Research map / Graph learning / Graph algorithms and clustering
Continuous DAG structure learning
Learn DAG structure via continuous optimization and active experimental design.
102 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Graph algorithms and clustering, keyword cluster “search · directed · causal”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
21 papers in 2023–24 and 27 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 10.7% to 13.4%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · Learning Granger Causality for Hawkes Processes (ICML 2016 · 74 citations)
- 2017 · Experimental Design for Learning Causal Graphs with Latent Variables (NeurIPS 2017 · 33 citations)
- 2018 · DAGs with NO TEARS: Continuous Optimization for Structure Learning (NeurIPS 2018 · 217 citations)
- 2019 · Deep imitation learning for molecular inverse problems (NeurIPS 2019 · 17 citations)
- 2020 · Global Greedy Dependency Parsing (AAAI 2020 · 35 citations)
- 2021 · Graphical Models in Heavy-Tailed Markets (NeurIPS 2021 · 12 citations)
- 2022 · Deep Attentive Belief Propagation: Integrating Reasoning and Learning for Solving Constraint Optimization Problems (NeurIPS 2022 · 2 citations)
- 2023 · Deep Graph Representation Learning and Optimization for Influence Maximization (ICML 2023 · 78 citations)
- 2024 · Graph Bayesian Optimization for Multiplex Influence Maximization (AAAI 2024 · 12 citations)
- 2025 · DiffIM: Differentiable Influence Minimization with Surrogate Modeling and Continuous Relaxation (AAAI 2025 · 0 citations)
- 2026 · BoGrape: Bayesian optimization over graphs with shortest-path encoded (ICLR 2026 · 1 citation)
Earliest papers here
- Learning Granger Causality for Hawkes Processes (ICML 2016 · 74 citations)
- Generating CP-Nets Uniformly at Random (AAAI 2016 · 13 citations)
- Experimental Design for Learning Causal Graphs with Latent Variables (NeurIPS 2017 · 33 citations)
What has been tried
70 papers here report 555 measured changes (a component added, removed or swapped, compared with a baseline): helped 66%, no clear effect 11%, hurt 18%, mixed 6%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| omit disjointness axioms | ORE 2015-derived test datasets · computation time | more than half of the computation time | Helped | Efficient Axiomatization of OWL 2 EL Ontologies from Data by Means of Formal Concept Analysis (AAAI 2024 · 2 citations) |
| Use content-based hashing without delta compression | G1–G5 · accuracy delta | 0.00 | No clear effect | MGit: A Model Versioning and Management System (ICML 2024 · 1 citation) |
| add the budget loss | WannaCry (WC), b=5 · reduced ratio of influence | 0.0052 -> 0.3914 | Helped | DiffIM: Differentiable Influence Minimization with Surrogate Modeling and Continuous Relaxation (AAAI 2025 · 0 citations) |
| Increase the input maximum degree estimate to at least 7 while using an infection probability estimate near the true value | chn.2009.flu.1.00, augmented graphs · F1-score | approximately 0.2 to 0.4; ≥ 7 | Helped | Learn to Vaccinate: Combining Structure Learning and Effective Vaccination for Epidemic and Outbreak Control (ICML 2025 · 0 citations) |
| Replace the soft-subset representation with a neural network representation | N=200, 800 edges · skeleton SHD | 81±9 -> 173±11 | Hurt | Scalable and Flexible Causal Discovery with an Efficient Test for Adjacency (ICML 2024 · 0 citations) |
| Select intervention pairs using random cost instead of min cost | K=2 simulation, PBM · MCC | Mixed | Identifiability Matters: Revealing the Hidden Recoverable Condition in Unbiased Learning to Rank (ICML 2024 · 0 citations) | |
| Increase the number of observations from 10 to 100 or more | ER-10-1, N column · p_LMC | 0.1045; 0.0084; 0.0028; 0.002 | Helped | Toward Falsifying Causal Graphs Using a Permutation-Based Test (AAAI 2025) |
| vary RG depth D from 1 to 100 | Prompt Engineering · average shortest reliable prompt length | Mixed | Multi-Objective Hyperparameter Selection via Hypothesis Testing on Reliability Graphs (NeurIPS 2025) |
What reviewers will expect
Datasets and baselines named in the 15 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Baselines
- Random · 2 of 15 papers
- Greedy · 2 of 15 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 48 papers; from OpenAlex and the papers' affiliations).
- Google · 3 papers
- University of Science and Technology of China · 2 papers
- Peking University · 2 papers
- University of California, Irvine · 2 papers
Use it
Download BibTeX of 39 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Neural combinatorial optimization on graphs (159 papers)
- Graph comparison and random features (155 papers)
- Graph embedding and spectral sparsification (126 papers)
- Graph distance and retrieval (79 papers)
Elsewhere in the topic
- Neural and local graph clustering (81 papers)
- Deep learning for graph matching (42 papers)
- Optimization algorithms for graph matching (42 papers)
- Anchor-based multi-view clustering (31 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).