Research map / Learning theory and algorithms / Kernels and regression theory
Sub-optimality of the Naive Mean Field approximation for proportional high-dimensional Linear Regression
Jiaze Qiu
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 the NMF distribution to build 95% credible regions for Gaussian spike-and-slab prior in the proportional regime | Naive Mean Field approximation · Gaussian spike-and-slab prior, iid Gaussian design, n/p = alpha | average coverage rate vs 95% nominal | Hurt |
| build credible intervals from quantiles of B given the NMF mean estimate instead of NMF posterior quantiles | NMF-based credible regions · proportional asymptotics, pi = pi* | asymptotic coverage | Improved |
| use NMF log-normalizing constant instead of the true one | NMF free energy · Gaussian spike-and-slab prior | log-normalizing constant accuracy | Hurt |
| NMF in the regime p = o(n) instead of p = Theta(n) | NMF approximation for linear regression · high-dimensional linear regression | approximation quality | Improved |
| NMF in the setting where posterior mean estimation error is measured | NMF with spike-and-slab prior · mean square error | MSE | No clear effect |
| set q close to 1 in the spike-and-slab prior when solving the fixed point equation | fixed-point evaluation of theory · Gaussian spike-and-slab | fixed point convergence | Hurt |
| check theory by gradient descent on the NMF objective with n=8000 | asymptotic predictions · Gaussian spike-and-slab | agreement with simulation | Improved |
Closest papers by meaning
- Gaussian Mean Field Variational Inference can Overestimate Predictive Variance (ICML 2026 · similarity 0.86)
- Covariances, Robustness, and Variational Bayes (JMLR 2018 · similarity 0.86)
- The Art of BART: Minimax Optimality over Nonhomogeneous Smoothness in High Dimension (JMLR 2023 · similarity 0.86)
- On the detrimental effect of invariances in the likelihood for variational inference (NeurIPS 2022 · similarity 0.86)
- Non-Asymptotic Uncertainty Quantification in High-Dimensional Learning (NeurIPS 2024 · similarity 0.85)
- Drifting Fields are not Conservative (NeurIPS 2026 · 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.
