Research map / Learning theory and algorithms / Prediction and decision losses
Deep Neural Networks Constrained by Decision Rules
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 attention over rule latent representations (RCN+A vs RCN-A) (ablation) | RCN with RF ruleset · FordA | error rate: .327 -> .096 | Helped |
| add attention over rule latent representations (RCN+A vs RCN-A) (ablation) | RCN with RF ruleset · Average rank over 8 datasets | average rank: 5.563 -> 1.813 | Helped |
| use Eclat ruleset instead of RF ruleset (ablation) | RCN+A · 8 datasets | error rate: 2.438 -> 1.813 avg rank | No clear effect |
| constrain neural network to select decision rules (RCN+A vs SBRL) | CNN/Bi-LSTM · IMDB | error rate: .300 -> .134 | Helped |
| constrain neural network to select decision rules (RCN+A RF vs base network) | Bi-LSTM · IMDB | error rate: .136 -> .134 | No clear effect |
| ruleset optimization with lambda=1000 (ablation) | RCN+A RF · IMDB | error rate: .134 -> .119 | Helped |
| ruleset optimization with lambda=100 (ablation) | RCN+A RF · Strawberry | error rate: .038 -> .029 | Helped |
| ruleset optimization with lambda=100 (ablation) | RCN+A RF · Yelp | error rate: .054 -> .091 | Hurt |
| shrink ruleset limit to lambda=10 (ablation) | RCN+A RF · FordB | error rate: .120 -> .348 | Hurt |
| shrink ruleset limit to lambda=10 (ablation) | RCN+A RF · Yelp | error rate: .054 -> .187 | Hurt |
| raise prior weight of human-made rules to rho=10 (ablation) | RCN+A RF · IMDB | error rate: 0.134 -> 0.145 | Hurt |
| raise prior weight of human-made rules to rho=100 (ablation) | RCN+A RF · IMDB | error rate: 0.134 -> 0.23 | Hurt |
Closest papers by meaning
- Explainable Neural Networks with Guarantee: A Sparse Estimation Approach (AAAI 2025 · similarity 0.84)
- Explaining, Fast and Slow: Abstraction and Refinement of Provable Explanations (ICML 2025 · similarity 0.84)
- Towards Modeling Uncertainties of Self-Explaining Neural Networks via Conformal Prediction (AAAI 2024 · similarity 0.83)
- HyperLogic: Enhancing Diversity and Accuracy in Rule Learning with HyperNets (NeurIPS 2024 · similarity 0.83)
- Understanding Generalization through Decision Pattern Shift (NeurIPS 2026 · similarity 0.82)
- Neurons to Words: A Novel Method for Automated Neural Network Interpretability and Alignment (AAAI 2025 · similarity 0.82)
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.