Research map / Graph learning / Graph self-supervised learning
Graphite: Iterative Generative Modeling of Graphs
Aditya Grover, Aaron Zweig, Stefano Ermon
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use an iterative learned graph-refinement decoder instead of the inner-product decoder | Graphite-VAE · Erdos-Renyi, Ego, Regular, Geometric, Power Law, Barabasi-Albert | negative log-likelihood: 273.82 -> 270.22; 273.76 -> 270.70; 275.29 -> 266.54; 274.09 -> 269.71; 278.86 -> 263.92; 274.4 -> 268.73 | Improved |
| Use iterative graph refinement rather than an inner-product-only decoder | Link prediction, featureless citation networks · Cora, Citeseer, Pubmed | AUC: Cora: 90.1 -> 91.5; Citeseer: 92.0 -> 93.5; Pubmed: 92.3 -> 94.6 | Mixed |
| Use iterative graph refinement rather than an inner-product-only decoder | Link prediction with node features · Cora*, Citeseer*, Pubmed* | AUC: 94.1 -> 94.7; 96.7 -> 97.3; 95.5 -> 97.4 | Mixed |
| Combine the Graphite objective with supervised GCN training | GCN node classifier · Cora*, Citeseer*, Pubmed* | accuracy: 81.5 -> 82.1; 70.3 -> 71.0; 79.0 -> 79.3 | Improved |
| Train using edge subsampling for Monte Carlo objective evaluation | Graphite · Cora, Citeseer, Pubmed | AUC: 90.5 vs 91.5; 92.5 vs 93.5; 93.1 vs 94.6 | Hurt |
| Use an embedding skip connection combining the initial and refined embeddings | Graphite · Link prediction | AUC | Improved |
Later papers that compared with this method
2 later accepted papers report Graphite as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Deconvolutional Networks on Graph Data (NeurIPS 2021 · better on 5 of 5 shared results)
- Permutation-Invariant Variational Autoencoder for Graph-Level Representation Learning (NeurIPS 2021 · better on 3 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Permutation-Invariant Variational Autoencoder for Graph-Level Representation Learning (NeurIPS 2021 · similarity 0.89)
- D-VAE: A Variational Autoencoder for Directed Acyclic Graphs (NeurIPS 2019 · similarity 0.89)
- Variational Graph Recurrent Neural Networks (NeurIPS 2019 · similarity 0.89)
- Constrained Generation of Semantically Valid Graphs via Regularizing Variational Autoencoders (NeurIPS 2018 · similarity 0.88)
- The quest for the GRAph Level autoEncoder (GRALE) (NeurIPS 2025 · similarity 0.86)
- Generative Graph Pattern Machine (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
