Research map / Learning theory and algorithms / Prediction and decision losses
Exploring and Interacting with the Set of Good Sparse Generalized Additive Models
Chudi Zhong, Zhi Chen, Jiachang Liu, Margo Seltzer, Cynthia Rudin
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 |
|---|---|---|---|
| use blocking (Method 2) instead of re-optimization (Method 1) for a new support set | Rashomon set approximation · COMPAS K~=15 | time (s): 324.85 -> 4.84e-4 | Improved |
| use blocking (Method 2) instead of re-optimization (Method 1) for a new support set | Rashomon set approximation · MIMIC-II K~=35 | time (s): 1127.10 -> 5.89e-4 | Improved |
| use blocking (Method 2) instead of Method 1 | Rashomon set approximation · FICO K~=28 | precision ratio: 1.01 | No clear effect |
| shrink support set from K~=15 to 13 bins | Method 2 · COMPAS | volume ratio: 1.05 -> 1.20 | Hurt |
| shrink support set from 28 to 24 bins | Method 2 · FICO | volume ratio: 0.90 -> 1.29 | Hurt |
| optimize log-determinant convex objective instead of original ellipsoid objective (ablation) | Method 1 · COMPAS/FICO/MIMIC-II | precision ratio: 1.01 -> 1.00 | No clear effect |
| solve LP with sign constraint instead of MIP for VI+ (fix other coefs) | variable importance upper bound · Diabetes | time (s): 18.876 -> 4.780 | Improved |
| solve LP with sign constraint instead of MIP for VI+ (not fix other coefs) | variable importance upper bound · COMPAS | time (s): 46.782 -> 3.043 | Improved |
| use proposed ellipsoid method instead of EBM+sampling | Rashomon set approximation · FICO lambda_s=5e-4 | running time (s): 1169.73 -> 131.71 | Improved |
| use proposed ellipsoid method instead of logistic regression+sampling | Rashomon set approximation · COMPAS lambda_s=5e-4 | running time (s): 22.65 -> 94.92 | Hurt |
| use proposed ellipsoid method instead of logistic regression+sampling | Rashomon set approximation · MIMIC-II lambda_s=5e-4 | running time (s): 786.85 -> 437.87 | Improved |
| increase theta from 1.005L* to 1.1L* | ellipsoid scaling needed · Diabetes | scaling ratio: 1.081 -> 1.119 | Hurt |
Closest papers by meaning
- Diverse Representative Rashomon Sets for Sparse Generalized Additive Models (NeurIPS 2026 · similarity 0.92)
- SORTeD Rashomon Sets of Sparse Decision Trees: Anytime Enumeration (NeurIPS 2025 · similarity 0.86)
- ElliCE: Efficient and Provably Robust Algorithmic Recourse via the Rashomon Sets (NeurIPS 2025 · similarity 0.85)
- Dropout-Based Rashomon Set Exploration for Efficient Predictive Multiplicity Estimation (ICLR 2024 · similarity 0.84)
- Exploring the cloud of feature interaction scores in a Rashomon set (ICLR 2024 · similarity 0.84)
- Rashomon Sets for Prototypical-Part Networks: Editing Interpretable Models in Real-Time (CVPR 2025 · 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.
