Research map / Graph learning / Knowledge and temporal graphs
Learning Graph Structure With A Finite-State Automaton Layer
Daniel Johnson, Hugo Larochelle, Daniel Tarlow
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 |
|---|---|---|---|
| Add GFSA edges halfway through a RAT | Variable misuse model · Variable misuse, All | full accuracy: 88.22 -> 89.63 | Improved |
| Add GFSA edges before a RAT | Variable misuse model · Variable misuse, All | full accuracy: 88.22 -> 89.47 | Improved |
| Add GFSA edges halfway through a RAT rather than before it | Variable misuse model · Variable misuse, All | full accuracy: 89.47 -> 89.63 | Improved |
| Add GFSA edges halfway through a RAT | Variable misuse model · Variable misuse, With bug classification | accuracy: 92.48 -> 94.25 | Improved |
| Add GFSA edges halfway through a RAT | Variable misuse model · Variable misuse, With bug Loc & Repair | accuracy: 88.39 -> 89.93 | Improved |
| Add GFSA edges before a RAT | Variable misuse model · Variable misuse, No bug classification | accuracy: 93.03 -> 93.10 | No clear effect |
| Add random-walk edges before a RAT | Variable misuse model · Variable misuse, No bug classification | accuracy: 92.05 -> 93.22 | Improved |
| Replace exact trajectory marginals and cross-entropy with single sampled edges trained using REINFORCE (ablation) | GFSA program-analysis edge classifier · Last Read, 100 training examples, 1x | F1: 98.81 -> 98.48 | Hurt |
| Replace exact trajectory marginals and cross-entropy with single sampled edges trained using REINFORCE (ablation) | GFSA program-analysis edge classifier · Last Write, 100 training examples, 1x | F1: 98.71 -> 98.32 | Hurt |
| Replace learned multi-step GFSA policies with a uniform random walk and learned halting probability (ablation) | Variable misuse model · Variable misuse, All | full accuracy: 88.22 -> 88.91 | No clear effect |
| Train learned GFSA navigation options | Grid-world pathfinding · Validation layouts | steps to goal: 40.1 -> 11.5 steps | Improved |
Closest papers by meaning
- Learning to Represent Programs with Graphs (ICLR 2018 · similarity 0.87)
- VQGraph: Rethinking Graph Representation Space for Bridging GNNs and MLPs (ICLR 2024 · similarity 0.86)
- Towards Graph Foundation Models: Learning Generalities Across Graphs via Task-Trees (ICML 2025 · similarity 0.85)
- AutoGFM: Automated Graph Foundation Model with Adaptive Architecture Customization (ICML 2025 · similarity 0.85)
- Generative Graph Pattern Machine (NeurIPS 2025 · similarity 0.84)
- GraphChain: Large Language Models for Large-scale Graph Analysis via Tool Chaining (NeurIPS 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.
