Research map / Probabilistic and scientific modeling / Molecules and proteins
Learning Approximate Inference Networks for Structured Prediction
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 |
|---|---|---|---|
| Replace gradient-based inference with a trained inference network | SPEN on Bookmarks · Bookmarks test | example-averaged F1: 34.4 -> 37.6 | Helped |
| Replace gradient-based inference with a trained inference network | SPEN on multi-label classification · Multi-label classification average | example-averaged F1: 38.0 -> 39.1 | Helped |
| Replace gradient-based inference with a trained inference network | SPEN on Twitter POS · Twitter POS test | accuracy: 89.2 -> 89.8 | Helped |
| Add tag language model energy (ablation) | SPEN with margin-rescaled hinge on Twitter POS · Twitter POS test | accuracy: 89.6 -> 90.2 | Helped |
| Use slack-rescaled hinge instead of contrastive hinge (ablation) | SPEN on Twitter POS · Twitter POS validation | accuracy: 89.0 -> 89.6 | Helped |
| Add retuning of the inference network after training (ablation) | SPEN with slack-rescaled hinge on Twitter POS · Twitter POS validation | accuracy: 89.4 -> 89.6 | Helped |
| Replace cross-entropy stabilization with entropy stabilization (ablation) | SPEN inference network on Twitter POS · Twitter POS validation | accuracy: 89.3 -> 86.8 | Hurt |
| Add cross-entropy stabilization to CRF inference network (ablation) | Trained CRF · Twitter POS test | accuracy: 89.2 -> 89.5 | Helped |
| Remove all stabilization terms from the CRF inference network (ablation) | Trained CRF · Twitter POS development | accuracy: 13.7% | Hurt |
| Add cross-entropy stabilization to the SPEN inference network (ablation) | SPEN on Twitter POS · Twitter POS validation | accuracy: 84.2 -> 89.1 | Helped |
| Add TLM energy with slack-rescaled hinge (ablation) | SPEN on Twitter POS · Twitter POS test | accuracy | Hurt |
| Use inference network instead of exact Viterbi decoding | CRF on Twitter POS · Twitter POS test | accuracy: 89.2 -> 89.5 | Helped |
Closest papers by meaning
- Structured Prediction Energy Networks (ICML 2016 · similarity 0.90)
- Search-Guided, Lightly-Supervised Training of Structured Prediction Energy Networks (NeurIPS 2019 · similarity 0.90)
- Graph Structured Prediction Energy Networks (NeurIPS 2019 · similarity 0.88)
- Structured Energy Network As a Loss (NeurIPS 2022 · similarity 0.88)
- End-to-End Learning for Structured Prediction Energy Networks (ICML 2017 · similarity 0.87)
- Gradient-Based Inference for Networks with Output Constraints (AAAI 2019 · 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.