Research map / Graph learning / Graph self-supervised learning
Junction Tree Variational Autoencoder for Molecular Graph Generation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Generate molecules in two phases using a junction-tree scaffold of valid substructures followed by graph assembly | Molecular graph generation on ZINC · ZINC prior samples | Validity: 89.2% -> 100.0% | Helped |
| Use a junction-tree and graph latent representation with staged decoding | Molecular reconstruction · ZINC test | Reconstruction: 76.2% -> 76.7% | Helped |
| Use JT-VAE latent embeddings for Bayesian optimization | Molecular property optimization · Bayesian optimization | Best property score: 4.04 -> 5.30 | Helped |
| Use JT-VAE latent embeddings rather than SD-VAE embeddings | Sparse Gaussian Process property prediction · ZINC, 10-fold cross validation | RMSE: 1.366 ± 0.023 -> 1.290 ± 0.026 | Helped |
| Tighten the similarity constraint from δ=0.0 to δ=0.6 | JT-VAE constrained optimization · Constrained optimization | Similarity: 0.28 ± 0.15 -> 0.69 ± 0.06 | Helped |
| Tighten the similarity constraint from δ=0.0 to δ=0.6 | JT-VAE constrained optimization · Constrained optimization | Improvement: 1.91 ± 2.04 -> 0.21 ± 0.71 | Hurt |
| Tighten the similarity constraint from δ=0.0 to δ=0.6 | JT-VAE constrained optimization · Constrained optimization | Success: 97.5% -> 46.4% | Hurt |
Closest papers by meaning
- Constrained Graph Variational Autoencoders for Molecule Design (NeurIPS 2018 · similarity 0.94)
- NeVAE: A Deep Generative Model for Molecular Graphs (AAAI 2019 · similarity 0.92)
- Spanning Tree-based Graph Generation for Molecules (ICLR 2022 · similarity 0.91)
- Lift Your Molecules: Molecular Graph Generation in Latent Euclidean Space (ICLR 2025 · similarity 0.87)
- Graph Generative Pre-trained Transformer (ICML 2025 · similarity 0.87)
- Bridging Molecular Graphs and Large Language Models (AAAI 2025 · similarity 0.87)
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.