Research map / Probabilistic and scientific modeling / Molecules and proteins
Syntax-Directed Variational Autoencoder for Structured Data
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 |
|---|---|---|---|
| Add online semantic constraints using stochastic lazy attributes | SD-VAE · Programs prior | valid prior (%): 2.96 -> 100.00 | Helped |
| Add online semantic constraints using stochastic lazy attributes | SD-VAE · Zinc SMILES prior | valid prior (%): 7.2 -> 43.5 | Helped |
| Add online semantic constraints using stochastic lazy attributes | SD-VAE · Programs test | reconstruction accuracy (%): 71.83 -> 96.46 | Helped |
| Add online semantic constraints using stochastic lazy attributes | SD-VAE · Zinc SMILES test | reconstruction accuracy (%): 53.7 -> 76.2 | Helped |
| Use semantic-guided decoding rather than CFG-only decoding | VAE latent representations on Programs · Programs test | RMSE: 3.378 ± 0.020 -> 3.185 ± 0.025 | Helped |
| Use semantic-guided decoding rather than CFG-only decoding | VAE latent representations on Zinc · Zinc test | RMSE: 1.404 ± 0.006 -> 1.366 ± 0.023 | Helped |
| Constrain molecule generation with semantic attributes | SD-VAE · Zinc generated molecules | PairFp diversity: 0.94 ± 0.10 -> 0.95 ± 0.08 | No clear effect |
| Use kekulized SMILES instead of the reported aromatic form | SD-VAE · Zinc SMILES prior | valid prior (%): 43.5 -> 97.3 | Helped |
| Use SD-VAE for Bayesian optimization rather than GVAE | Program optimization · Programs | distance to ground truth | Helped |
Closest papers by meaning
- Grammar Variational Autoencoder (ICML 2017 · similarity 0.92)
- From Variational to Deterministic Autoencoders (ICLR 2020 · similarity 0.88)
- Discrete Object Generation with Reversible Inductive Construction (NeurIPS 2019 · similarity 0.86)
- Constrained Graph Variational Autoencoders for Molecule Design (NeurIPS 2018 · similarity 0.86)
- Controlled Generation with Equivariant Variational Flow Matching (ICML 2025 · similarity 0.85)
- Syntactic and Semantic Control of Large Language Models via Sequential Monte Carlo (ICLR 2025 · similarity 0.84)
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.