Research map / Label-efficient and robust learning / Recognition and long-tail
A Performance-Driven Benchmark for Feature Selection in Tabular Deep Learning
Valeriia Cherepanova, Roman Levin, Gowthami Somepalli, Jonas Geiping, C. Bayan Bruss, Andrew Gordon Wilson, Tom Goldstein, Micah Goldblum
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 |
|---|---|---|---|
| select features with Deep Lasso instead of training on all features | MLP · random extra features, 50% | average rank: 8.08 -> 2.33 | Improved |
| select features with Deep Lasso instead of training on all features | MLP · corrupted extra features, 75% | average rank: 7.58 -> 1.42 | Improved |
| select features with Deep Lasso instead of training on all features | MLP · second-order extra features, 50% | average rank: 4.50 -> 2.67 | Improved |
| select features with Deep Lasso instead of training on all features | MLP · second-order extra features, 75% | average rank: 4.92 -> 2.17 | Improved |
| select features with Deep Lasso instead of training on all features | FT-Transformer · random extra features, 50% | average rank: 6.58 -> 3.66 | Improved |
| select features with Deep Lasso instead of XGBoost importance | MLP · corrupted extra features, 75% | average rank: 2.58 -> 1.42 | Improved |
| select features with Deep Lasso instead of XGBoost importance | MLP · second-order extra features, 75% | average rank: 5.00 -> 2.17 | Improved |
| select features with XGBoost importance instead of Deep Lasso | FT-Transformer · random extra features, 50% | average rank: 3.66 -> 1.91 | Improved |
| select features with XGBoost importance instead of Deep Lasso | FT-Transformer · second-order extra features, 50% | average rank: 2.41 -> 4.08 | Hurt |
| use feature selection on second-order features for FT-Transformer with Univariate test | FT-Transformer · second-order extra features, 50% | average rank: 4.08 -> 6.25 | Hurt |
| apply Deep Lasso when 75% instead of 50% of features are second-order | MLP · second-order extra features | average rank: 2.67 -> 2.17 | Improved |
| add Deep Lasso feature selection to the MLP versus XGBoost baseline | MLP · random extra features, 50% | average rank: 2.58 -> 1.42 | Improved |
Closest papers by meaning
- TabReD: Analyzing Pitfalls and Filling the Gaps in Tabular Deep Learning Benchmarks (ICLR 2025 · similarity 0.88)
- TabArena: A Living Benchmark for Machine Learning on Tabular Data (NeurIPS 2025 · similarity 0.87)
- TabPrep: Closing the Feature Engineering Gap in Tabular Benchmarks (NeurIPS 2026 · similarity 0.87)
- A Data-Centric Perspective on Evaluating Machine Learning Models for Tabular Data (NeurIPS 2024 · similarity 0.86)
- TabularBench: Benchmarking Adversarial Robustness for Tabular Deep Learning in Real-world Use-cases (NeurIPS 2024 · similarity 0.86)
- TabR: Tabular Deep Learning Meets Nearest Neighbors (ICLR 2024 · 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.
