Research map / Learning theory and algorithms / Matrix and tensor methods
Spectral Deconfounding via Perturbed Sparse Linear Models
Domagoj Ćevid, Peter Bühlmann, Nicolai Meinshausen
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 Trim or Lava spectral transformation before the Lasso | Lasso · simulated confounding model (n=200, p=600, q=6) | l1 estimation error of beta | Improved |
| add Trim or Lava spectral transformation before the Lasso when there is no confounding (b=0) | Lasso · simulated sparse linear model with spiked covariance | l1 estimation error of beta | No clear effect |
| use Puffer transform instead of Trim when n and p are close | spectral transformation before Lasso · simulated confounding model | l1 estimation error of beta | Hurt |
| misspecify the number of removed principal components in PCA adjustment (ablation) | PCA adjustment · simulated confounding model | l1 estimation error of beta | Hurt |
| choose lambda by cross-validation instead of oracle value (ablation) | Puffer transform · simulated confounding model | l1 estimation error of beta | Hurt |
| choose lambda by cross-validation instead of oracle value (ablation) | Trim transform and Lava · simulated confounding model | l1 estimation error of beta | No clear effect |
| randomize the direction of the perturbation b (perturbed linear model) instead of confounding direction (ablation) | standard Lasso · simulated data | l1 estimation error of beta | Improved |
| add Trim transform or Lava before the Lasso | Lasso · GTEx skeletal muscle gene expression (p=14713, n=491) | Jaccard distance of supports (original vs deconfounded data) | Improved |
Closest papers by meaning
- High-Dimensional Inference for Generalized Linear Models with Hidden Confounding (JMLR 2023 · similarity 0.87)
- The Complete Lasso Tradeoff Diagram (NeurIPS 2020 · similarity 0.87)
- Feature Adaptation for Sparse Linear Regression (NeurIPS 2023 · similarity 0.86)
- Free Decompression with Algebraic Spectral Curves (NeurIPS 2026 · similarity 0.85)
- Spectral Estimation with Deformed Decompression (NeurIPS 2026 · similarity 0.85)
- Spectral Estimation with Free Decompression (NeurIPS 2025 · similarity 0.85)
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.
