Research map / Label-efficient and robust learning / Recognition and long-tail
Unveiling the Role of Data Uncertainty in Tabular Deep Learning
Nikolay Kartashev, Ivan Rubachev, Artem Babenko
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 |
|---|---|---|---|
| add learned LRLR numerical embeddings | MLP · CIFAR-10 (tabular, original) | accuracy: 58.31 -> 57.19 | Hurt |
| add learned LRLR numerical embeddings | MLP · Uncertain CIFAR-10 | accuracy: 47.54 -> 48.73 | Improved |
| replace MLP with retrieval-based ModernNCA | MLP · Uncertain CIFAR-10 | accuracy: 47.54 -> 48.45 | Improved |
| replace deep ensemble with TabM | Deep Ensemble · CIFAR-10 (tabular, original) | accuracy: 60.92 -> 61.01 | No clear effect |
| replace deep ensemble with TabM | Deep Ensemble · Uncertain CIFAR-10 | accuracy: 49.41 -> 50.45 | Improved |
| pretrain LRLR embeddings with triplet loss on target similarity | MLP-LRLR · regression benchmark | average rank: 1.67 -> 1.17 | Improved |
| pretrain LRLR embeddings with triplet loss on target similarity | MLP-LRLR · regression benchmark | change vs MLP (%): 2.00 -> 2.81 | Improved |
| add periodic numerical embeddings (PLR) | MLP · Delivery ETA | MSE | Mixed |
| use ModernNCA retrieval head | MLP · Maps Routing and House (low/medium uncertainty) | MSE | Hurt |
| average gradients across TabM branches (ablation) | TabM · synthetic MLP-parameterized dataset | clean-to-noisy gradient ratio | Improved |
| independent deep ensemble of MLPs instead of shared-parameter ensemble | MLP · saw-like 2D synthetic dataset | prediction quality | No clear effect |
Closest papers by meaning
- TabReD: Analyzing Pitfalls and Filling the Gaps in Tabular Deep Learning Benchmarks (ICLR 2025 · similarity 0.89)
- TabM: Advancing tabular deep learning with parameter-efficient ensembling (ICLR 2025 · similarity 0.89)
- TabR: Tabular Deep Learning Meets Nearest Neighbors (ICLR 2024 · similarity 0.88)
- TabularBench: Benchmarking Adversarial Robustness for Tabular Deep Learning in Real-world Use-cases (NeurIPS 2024 · similarity 0.88)
- TabSTAR: A Tabular Foundation Model for Tabular Data with Text Fields (NeurIPS 2025 · similarity 0.86)
- TabArena: A Living Benchmark for Machine Learning on Tabular Data (NeurIPS 2025 · similarity 0.86)
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.
