Research map / Architectures and efficiency / RNNs and normalization
An Inductive Bias for Tabular Deep Learning
Ege Beyazit, Jonathan Kozaczuk, Bo Li, Vanessa Wallace, Bilal H Fadlallah
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 |
|---|---|---|---|
| replace unit-length scaling with rank transform | MLP · MiniBooNE | accuracy: 90.78 ± 0.04 -> 95.01 ± 0.06 | Improved |
| replace unit-length scaling with standardization | MLP · MiniBooNE | accuracy: 90.78 ± 0.04 -> 94.0 ± 0.04 | Improved |
| replace unit-length scaling with rank transform | MLP · bank-marketing | accuracy: 90.56 ± 0.11 -> 90.41 ± 0.08 | No clear effect |
| replace unit-length scaling with selrank | MLP · electricity | accuracy: 85.28 ± 0.12 -> 87.04 ± 0.17 | Improved |
| replace unit-length scaling with scale | MLP · electricity | accuracy: 85.28 ± 0.12 -> 85.19 ± 0.13 | No clear effect |
| replace unit-length scaling with rank | MLP · albert | accuracy: 66.35 ± 0.04 -> 66.21 ± 0.08 | No clear effect |
| replace unit-length scaling with selrank | MLP · california | accuracy: 88.51 ± 0.15 -> 88.48 ± 0.2 | No clear effect |
| scale instead of unit length reduces high-frequency energy | MLP · MiniBooNE | normalized high-frequency amplitude: 16.24 -> 0.35 | Improved |
| rank instead of unit length reduces high-frequency energy | MLP · house_16H | normalized high-frequency amplitude: 1.31 -> 0.11 | Improved |
| selrank vs rank alone | MLP · bank-marketing | normalized high-frequency amplitude: 0.74 -> 1.28 | Hurt |
| feed proposed layer's low-frequency features to TabNet | TabNet · 14 tabular datasets | AUROC: 4% ± 1.6% | Improved |
| apply proposed approach to XGB | XGB · 14 tabular datasets | performance | No clear effect |
Closest papers by meaning
- Hybrid Autoencoders for Tabular Data: Leveraging Model-Based Augmentation in Low-Label Settings (NeurIPS 2025 · similarity 0.88)
- Arithmetic Feature Interaction Is Necessary for Deep Tabular Learning (AAAI 2024 · similarity 0.87)
- TabM: Advancing tabular deep learning with parameter-efficient ensembling (ICLR 2025 · similarity 0.87)
- TabReD: Analyzing Pitfalls and Filling the Gaps in Tabular Deep Learning Benchmarks (ICLR 2025 · similarity 0.86)
- Retrieval & Fine-Tuning for In-Context Tabular Models (NeurIPS 2024 · similarity 0.85)
- TabR: Tabular Deep Learning Meets Nearest Neighbors (ICLR 2024 · 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.
