Research map / Probabilistic and scientific modeling / Variational and generative sampling
Normalizing Flows for Knockoff-free Controlled Feature Selection
Derek Hansen, Brian Manzo, Jeffrey Regier
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 FlowSelect instead of DDLK | controlled feature selection · scRNA-seq runtime | median runtime (min): 91.9 -> 59.4 | Improved |
| use FlowSelect instead of DeepKnockoff | controlled feature selection · scRNA-seq runtime | median runtime (min): 3.0 -> 59.4 | Hurt |
| use FlowSelect instead of KnockoffGAN | controlled feature selection · scRNA-seq runtime | median runtime (min): 3.73 -> 59.4 | Hurt |
| use FlowSelect instead of MASS | controlled feature selection · scRNA-seq runtime | median runtime (min): 12.6 -> 59.4 | Hurt |
| replace deep knockoff generators (DeepKnockoff, KnockoffGAN, MASS, DDLK) with FlowSelect conditional randomization test | FlowSelect · Mixture-of-Gaussians and scRNA-seq, targets 0.05/0.1/0.25 | FDR control | Improved |
| replace knockoff methods that do control FDR with FlowSelect | FlowSelect · Mixture-of-Gaussians and scRNA-seq | power | Improved |
| give MASS the true data distribution | MASS · Mixture-of-Gaussians | FDR control | No clear effect |
| use oracle Model-X knockoffs with true parameters versus FlowSelect | Oracle Model-X · Mixture-of-Gaussians | power | Hurt |
| give DDLK the exact joint density | DDLK · Mixture-of-Gaussians | FDR and power | No clear effect |
| replace normalizing flow with mixture density networks (HRT) (ablation) | FlowSelect · Mixture-of-Gaussians | FDR control | Hurt |
| replace normalizing flow with mixture density networks (HRT) (ablation) | FlowSelect · scRNA-seq | power and FDR | No clear effect |
| increase number of MCMC samples (ablation) | FlowSelect · scRNA-seq | power | Improved |
Closest papers by meaning
- DeepDRK: Deep Dependency Regularized Knockoff for Feature Selection (NeurIPS 2024 · similarity 0.84)
- Statistical Test for Feature Selection Pipelines by Selective Inference (ICML 2025 · similarity 0.82)
- Knockoffs Inference for Partially Linear Models with Automatic Structure Discovery (AAAI 2025 · similarity 0.82)
- Flow Density Control: Generative Optimization Beyond Entropy-Regularized Fine-Tuning (NeurIPS 2025 · similarity 0.81)
- Distributionally Robust Feature Selection (NeurIPS 2025 · similarity 0.80)
- Multivariate Conformal Selection (ICML 2025 · similarity 0.80)
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.
