Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Bayesian subset selection and variable importance for interpretable prediction and classification
Daniel R. Kowal
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 smallest acceptable subset S_small instead of S_min for selection | bbound(bayes) · simulation n=50, p=50, SNR=0.25 | TNR: 0.81 -> 0.98 | Improved |
| use S_small instead of S_min | bbound(bayes) · simulation n=500, p=50, SNR=1 | TNR: 0.67 -> 0.99 | Improved |
| use S_small instead of S_min | bbound(bayes) · simulation n=500, p=50, SNR=0.25 | TPR: 0.99 -> 0.86 | Hurt |
| S_small versus adaptive lasso for variable selection | adaptive lasso · simulation n=200, p=400, SNR=1 | TPR: 0.97 -> 0.93 | No clear effect |
| S_small versus adaptive lasso at small n, high SNR | adaptive lasso · simulation n=50, p=50, SNR=1 | TPR: 0.54 -> 0.34 | Hurt |
| posterior HPD selection instead of S_small | S_small · simulation n=200, p=400, SNR=1 | TPR: 0.93 -> 0.75 | Hurt |
| forward selection instead of S_small | S_small · simulation n=50, p=50, SNR=0.25 | TNR: 0.98 -> 0.06 | Hurt |
| classical subset selection instead of S_small | S_small · simulation n=500, p=50, SNR=1 | TNR: 0.99 -> 0.83 | Hurt |
| use branch-and-bound search instead of lasso/forward/backward search paths | bbound(bayes) · simulation, acceptable family size | number of acceptable subsets: up to 10 times the number of acceptable subsets | Improved |
| branch-and-bound search on NC education out-of-sample splits | bbound(bayes) · NC education 20 splits | number of subsets: 10-25 times the number of subsets | Improved |
| reduce m_k from 100 to 15 (ablation) | bbound(bayes) · NC education reading scores | number of acceptable subsets: 1183 -> 197 | No clear effect |
| increase epsilon from 0.1 to 0.2 (ablation) | bbound(bayes) · NC education reading scores | number of acceptable subsets: 1183 -> 977 | No clear effect |
Closest papers by meaning
- Leveraging Importance Weights in Subset Selection (ICLR 2023 · similarity 0.83)
- Towards a Unified Framework for Uncertainty-aware Nonlinear Variable Selection with Theoretical Guarantees (NeurIPS 2022 · similarity 0.81)
- Variable Importance Using Decision Trees (NeurIPS 2017 · similarity 0.81)
- Estimating Conditional Mutual Information for Dynamic Feature Selection (ICLR 2024 · similarity 0.80)
- Most Influential Subset Selection: Challenges, Promises, and Beyond (NeurIPS 2024 · similarity 0.79)
- Contextual Feature Selection with Conditional Stochastic Gates (ICML 2024 · similarity 0.79)
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.
