Research map / Architectures and efficiency / RNNs and normalization
Neural Oblivious Decision Ensembles for Deep Learning on Tabular Data
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 |
|---|---|---|---|
| Use NODE instead of the strongest default GBDT baseline | NODE · Epsilon, YearPrediction, Higgs, Microsoft, Yahoo, Click | classification error or MSE: 0.1119 -> 0.1043; 80.68 -> 77.43; 0.2434 -> 0.2412; 0.5587 -> 0.5584; 0.5756 -> 0.5666; 0.3438 -> 0.3309 | Helped |
| Use NODE instead of the strongest tuned baseline | NODE · Epsilon, YearPrediction, Higgs, Click | classification error or MSE: 0.1041 -> 0.1034; 78.53 -> 76.21; 0.2140 -> 0.2101; 0.3334 -> 0.3312 | Helped |
| Use NODE instead of tuned XGBoost | NODE · Microsoft and Yahoo | MSE: 0.5544 -> 0.5570; 0.5420 -> 0.5692 | Hurt |
| Replace entmax with softmax (ablation) | NODE, 1-layer YearPrediction · YearPrediction | MSE: 77.43 -> 78.41 | Hurt |
| Replace entmax with Gumbel-Softmax (ablation) | NODE, 1-layer YearPrediction · YearPrediction | MSE: 77.43 -> 79.39 | Hurt |
| Replace entmax with sparsemax (ablation) | NODE, 1-layer YearPrediction · YearPrediction | MSE: 77.43 -> 78.13 | Hurt |
| Replace entmax with softmax (ablation) | NODE, 4-layer YearPrediction · YearPrediction | MSE: 76.21 -> 77.58 | Hurt |
| Replace entmax with Gumbel-Softmax (ablation) | NODE, 4-layer YearPrediction · YearPrediction | MSE: 76.21 -> 79.69 | Hurt |
| Replace entmax with sparsemax (ablation) | NODE, 4-layer YearPrediction · YearPrediction | MSE: 76.21 -> 76.60 | Helped |
| Replace entmax with softmax (ablation) | NODE, 4-layer Epsilon · Epsilon | classification error: 0.1033 -> 0.1034 | No clear effect |
| Replace entmax with Gumbel-Softmax (ablation) | NODE, 4-layer Epsilon · Epsilon | classification error: 0.2908 -> 0.1033 | Hurt |
| Replace entmax with sparsemax (ablation) | NODE, 4-layer Epsilon · Epsilon | classification error: 0.1058 -> 0.1033 | Hurt |
Closest papers by meaning
- GRANDE: Gradient-Based Decision Tree Ensembles for Tabular Data (ICLR 2024 · similarity 0.89)
- DOFEN: Deep Oblivious Forest ENsemble (NeurIPS 2024 · similarity 0.88)
- TabM: Advancing tabular deep learning with parameter-efficient ensembling (ICLR 2025 · similarity 0.87)
- TabR: Tabular Deep Learning Meets Nearest Neighbors (ICLR 2024 · similarity 0.87)
- Hybrid Autoencoders for Tabular Data: Leveraging Model-Based Augmentation in Low-Label Settings (NeurIPS 2025 · similarity 0.86)
- Revisiting Nearest Neighbor for Tabular Data: A Deep Tabular Baseline Two Decades Later (ICLR 2025 · similarity 0.85)
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.