Research map / Learning theory and algorithms / Prediction and decision losses
Forward-Backward Selection with Early Dropping
Giorgos Borboudakis, Ioannis Tsamardinos
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 |
|---|---|---|---|
| Use early dropping with K=0 in forward-backward selection | FBS · Null data, p=200, alpha=0.05 | number of selected variables (type I errors): 20.8 -> 5.6 | Improved |
| Use early dropping with K=1 in forward-backward selection | FBS · Null data, p=200, alpha=0.1 | number of selected variables (type I errors): 38.1 -> 15.9 | Improved |
| Use early dropping with K=infinity | FBS · Null data, p=200, alpha=0.05 | number of selected variables (type I errors): 20.8 -> 22.3 | No clear effect |
| Use early dropping with K=0 or 1 | FBS · 12 binary datasets, same hyper-parameters | number of independence tests: around 1-2 orders of magnitude fewer tests | Improved |
| Use early dropping with K=infinity | FBS · 12 binary datasets, same hyper-parameters | number of independence tests: 25%-30% of the tests | Improved |
| Use early dropping with K=0 or 1 | FBS · 12 binary datasets, same hyper-parameters | AUC / accuracy | No clear effect |
| Use early dropping with K=0 (IT) | FBS (IT) · musk | AUC: 96.9 -> 93.0 | Hurt |
| Use early dropping with K=0 (EBIC) | FBS (EBIC) · 12 binary datasets, optimized hyper-parameters | number of selected variables: avg rank 3.21 -> 1.17 | Improved |
| Use independence test instead of EBIC | FBS-ED K=0 · 12 binary datasets, optimized hyper-parameters | AUC average rank: 6.33 -> 3.75 | Improved |
| Use LASSO-FS instead of FBS-ED K=0 (IT) | FBS-ED K=0 (IT) · 12 binary datasets, optimized hyper-parameters | AUC average rank: 3.75 -> 1.42 | Improved |
| Use LASSO-FS instead of FBS-ED K=0 (IT) | FBS-ED K=0 (IT) · arcene | AUC: 78.6 -> 83.7 | Improved |
| Limit LASSO-FS to the same number of variables as forward selection | LASSO-FS · 12 binary datasets | AUC / accuracy difference | No clear effect |
Closest papers by meaning
- A Comparison of Hamming Errors of Representative Variable Selection Methods (ICLR 2022 · similarity 0.88)
- Distributionally Robust Feature Selection (NeurIPS 2025 · similarity 0.85)
- Estimating Conditional Mutual Information for Dynamic Feature Selection (ICLR 2024 · similarity 0.85)
- SAND: One-Shot Feature Selection with Additive Noise Distortion (ICML 2025 · similarity 0.85)
- Best Subset Selection: Optimal Pursuit for Feature Selection and Elimination (ICML 2025 · similarity 0.84)
- Differentially Private High-dimensional Variable Selection via Integer Programming (NeurIPS 2025 · similarity 0.84)
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.
