Research map / Learning theory and algorithms / Prediction and decision losses
TT-Sparse: Learning Sparse Rule Models with Differentiable Truth Tables
Hans Farrell Soegeng, Sarthak Ketanbhai Modi, Thomas Peyrin
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 |
|---|---|---|---|
| learn fan-in with soft TopK mapping instead of FC + L0 gates (ablation) | TT-Sparse · Heart | AUC: .893 -> .931 | Improved |
| learn fan-in with soft TopK mapping instead of FC + L1 + prune (ablation) | TT-Sparse · Diabetes | AUC: .676 -> .829 | Improved |
| learn connectivity with W_map instead of magnitude TopK (ablation) | TT-Sparse · Diabetes | AUC: .716 -> .829 | Improved |
| use W_map instead of magnitude TopK (ablation) | TT-Sparse · Abalone | R2: .152 -> .512 | Improved |
| use soft TopK mapping instead of FC + L0 gates (ablation) | TT-Sparse · Abalone | R2: -.00 -> .512 | Improved |
| use soft TopK mapping instead of FC + L1 + prune (ablation) | TT-Sparse · Yeast | rule complexity: 96 -> 110 | Hurt |
| use closed-form sigmoid soft TopK instead of Sinkhorn | TT-Sparse · Differentiable TopK benchmark | memory (MB): 1{,}075.7 -> 3.4 | Improved |
| use closed-form soft TopK instead of perturbed maximizer | TT-Sparse · Differentiable TopK benchmark | gradient variance: 200.3 / 22.1 -> 0.0 | Improved |
| add don't-care terms to QMC minimization (k=6) (ablation) | TT-Sparse · per-node rule complexity, 28 datasets | literals per node: 19.24 -> 7.13 | Improved |
| use soft TopK instead of slot-based softmax selection from DWN (ablation) | TT-Sparse · 28 datasets | AUC/R2 | Improved |
| add skip connection from input features to classifier (ablation) | TT-Sparse · 28 datasets | AUC/R2 | Improved |
| increase thermometer bits beyond 7 (ablation) | TT-Sparse · 28 datasets | AUC/R2 | No clear effect |
Closest papers by meaning
- Explainable Neural Networks with Guarantee: A Sparse Estimation Approach (AAAI 2025 · similarity 0.88)
- Dynamic Sparse No Training: Training-Free Fine-tuning for Sparse LLMs (ICLR 2024 · similarity 0.85)
- Random Sparse Lifts: Construction, Analysis and Convergence of finite sparse networks (ICLR 2024 · similarity 0.85)
- Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models (ICLR 2025 · similarity 0.84)
- On the Complexity of Learning Sparse Functions with Statistical and Gradient Queries (NeurIPS 2024 · similarity 0.84)
- Codebook Features: Sparse and Discrete Interpretability for Neural Networks (ICML 2024 · 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.
