Research map / Probabilistic and scientific modeling / Bayesian optimization
Re-Examining Linear Embeddings for High-Dimensional Bayesian Optimization
Ben Letham, Roberto Calandra, Akshara Rai, Eytan Bakshy
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 Mahalanobis kernel (instead of ARD Matern 5/2) (ablation) | ALEBO · Branin D=100 | best value (minimize) | Improved |
| add hypersphere-sampled projection matrix (instead of iid normal) (ablation) | ALEBO · Branin D=100 | best value (minimize) | Improved |
| use Mahalanobis kernel instead of ARD RBF kernel in the embedding (ablation) | GP in 6-d embedding · Hartmann6 D=100 test predictions | predictive accuracy | Improved |
| increase embedding dimension d_e from d to above d (d=6: d_e 6 -> 12) (ablation) | random linear embedding · probability embedding contains an optimum (D=100, d=6) | P_opt: nearly 0 -> 0.5 | Improved |
| sample projection rows from unit hypersphere instead of N(0,1) or HeSBO (ablation) | random linear embedding · probability embedding contains an optimum | P_opt | Improved |
| increase d_e from 2 to larger values (ablation) | ALEBO · Branin D=100 | final best value | Improved |
| increase ambient dimension D from 100 to 1000 (ablation) | ALEBO · Branin | final best value | No clear effect |
| use ALEBO instead of REMBO | REMBO · Hartmann6 D=1000 | best value (minimize) | Improved |
| use ALEBO instead of REMBO | REMBO · Constrained NAS D=36 | CIFAR10 test accuracy | Improved |
| use CMA-ES instead of ALEBO | ALEBO · Daisy locomotion D=72 | reward | Hurt |
Closest papers by meaning
- Scalable Bayesian Optimization via Focalized Sparse Gaussian Processes (NeurIPS 2024 · similarity 0.88)
- BOIDS: High-Dimensional Bayesian Optimization via Incumbent-Guided Direction Lines and Subspace Embeddings (AAAI 2025 · similarity 0.88)
- Joint Composite Latent Space Bayesian Optimization (ICML 2024 · similarity 0.86)
- Bayesian Optimization on Function Spaces via Sparse RKHS Manifolds (NeurIPS 2026 · similarity 0.86)
- Understanding High-Dimensional Bayesian Optimization (ICML 2025 · similarity 0.86)
- Posterior Inference with Diffusion Models for High-dimensional Black-box Optimization (ICML 2025 · similarity 0.86)
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.
