Research map / Learning theory and algorithms / Prediction and decision losses
Beyond Additive Decompositions: Interpretability Through Separability
Jinyang Liu, Munir Hiabu
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 |
|---|---|---|---|
| remove positivity constraint and ordered-difference structure (TSL 1-product, R<=4) vs TSL R<=2 (ablation) | TSL · socmob | RMSE: 10.58 -> 9.87 | Improved |
| add positivity and ordered-difference structure vs TSL 1-product (ablation) | TSL · naval propulsion plant | RMSE: 0.0027 -> 0.0013 | Improved |
| add positivity and ordered-difference structure vs TSL 1-product (ablation) | TSL · auction verification | RMSE: 1336.51 -> 1135.80 | Improved |
| TSL R<=10 vs additive EBM | TSL · socmob | RMSE: 20.21 -> 9.48 | Improved |
| TSL R<=2 vs SepALS (r<=2) with fixed basis, matched rank | TSL · california housing | RMSE: 62162.22 -> 49376.09 | Improved |
| TSL vs EBM on largely additive problems | TSL · red wine | RMSE: 0.5991 -> 0.6438 | Hurt |
| TSL vs SepALS | TSL · naval propulsion plant, forest fires, kin8nm, energy efficiency, grid stability | RMSE | Hurt |
| TSL R<=10 vs depth-2 interpretable XGBoost | TSL · brazilian houses | RMSE: 2528.95 -> 2398.68 | Improved |
| TSL R<=10 vs black-box XGBoost | TSL · miami housing | RMSE: 89692.96 vs 82325.09 | Hurt |
Closest papers by meaning
- Scalable Interpretability via Polynomials (NeurIPS 2022 · similarity 0.86)
- Provably Explaining Neural Additive Models (ICLR 2026 · similarity 0.85)
- Analyzing Tensor Power Method Dynamics in Overcomplete Regime (JMLR 2017 · similarity 0.84)
- SHAP Meets Tensor Networks: Provably Tractable Explanations with Parallelism (NeurIPS 2025 · similarity 0.84)
- Explainable Neural Networks with Guarantee: A Sparse Estimation Approach (AAAI 2025 · similarity 0.84)
- Generalized Tensor Decomposition for Understanding Multi-Output Regression under Combinatorial Shifts (NeurIPS 2024 · similarity 0.83)
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.
