Research map / Learning theory and algorithms / Prediction and decision losses
Fast and Flexible Monotonic Functions with Ensembles of Lattices
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 |
|---|---|---|---|
| impose monotonicity on 4 features of a single lattice (ablation) | single lattice · ADULT (Dataset 1) | test accuracy: 84.96 -> 85.36 | Helped |
| use ensemble of lattices (Crystals) instead of single monotonic lattice (ablation) | Monotonic Crystals · ADULT (Dataset 1) | test accuracy: 85.36 -> 85.53 | Helped |
| Crystals versus random forest | Monotonic Crystals · ADULT (Dataset 1) | test accuracy: 85.21 -> 85.53 | Helped |
| Crystals versus random forest on large Dataset 3 | Crystals · Dataset 3 | test accuracy: 0.25% - 0.30% | Helped |
| Crystals versus random forest on same-distribution Test Set 1 | Crystals · Dataset 4 Test Set 1 | accuracy: 75.23 -> 75.18 | No clear effect |
| Crystals versus random forest on uniformly sampled Test Set 2 | Crystals · Dataset 4 Test Set 2 | accuracy: 90.51 -> 91.15 | Helped |
| train lattices independently instead of jointly (ablation) | RTL ensemble · Dataset 4 Test Set 1 | accuracy: 74.78 -> 72.80 | Hurt |
| use shared calibrators instead of separate calibrators (ablation) | RTL ensemble · Dataset 4 Test Set 1 | accuracy: 74.78 -> 74.48 | Hurt |
| choose feature subsets with Crystals instead of random subsets (RTL) (ablation) | Crystals ensemble · Dataset 2 | normalized MSE: 1000× | Helped |
| merge 300 lattices into one equivalent lattice when calibrators are shared | lattice ensemble · 14-feature problem | memory and evaluation time: 153 600 -> 16 384 | Helped |
| separate calibrators versus shared calibrators on evaluation time | lattice ensemble · evaluation | evaluation time | Hurt |
| compare Crystals evaluation cost to random forest | Crystals · Experiment 3, D = 50 | evaluation time and memory: 10× | Helped |
Closest papers by meaning
- Monotonic Kronecker-Factored Lattice (ICLR 2021 · similarity 0.89)
- GRANDE: Gradient-Based Decision Tree Ensembles for Tabular Data (ICLR 2024 · similarity 0.83)
- Free Lunch in the Forest: Functionally-Identical Pruning of Boosted Tree Ensembles (AAAI 2025 · similarity 0.83)
- Smooth Min-Max Monotonic Networks (ICML 2024 · similarity 0.83)
- Learning Monotonic Probabilities with a Generative Cost Model (ICML 2025 · similarity 0.82)
- No Free Lunch from Random Feature Ensembles: Scaling Laws and Near-Optimality Conditions (ICML 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.