Research map / Probabilistic and scientific modeling / Bayesian optimization
Bayesian Optimization with Robust Bayesian Neural Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| scale adaptation of SGHMC (automatic mass matrix and friction/step-size choice) (ablation) | SGHMC for Bayesian neural network regression · UCI regression benchmarks (Boston, Yacht, Concrete, Wine Red) | log likelihood | Helped |
| per-dataset tuning of SGHMC step size and friction vs a single shared setting | vanilla SGHMC · UCI regression benchmarks | log likelihood | Mixed |
| BOHAMIANN (BNN-based BO) vs GP-based BO on common benchmarks | Bayesian optimization surrogate model · Synthetic and hyperparameter benchmarks (Eggensperger et al.) | regret | No clear effect |
| multi-task BO with BNN (MT-BOHAMIANN) vs single-task GP BO | Multi-task hyperparameter optimization of SVM/RF · 21 OpenML datasets, 4 groups | speed to reach immediate regret <=0.25: 12% faster than GP BO (MTBO 5% faster) | Helped |
| BO-tuned hyperparameters of DDPG (BOHAMIANN) vs original DDPG | DDPG on CartPole · CartPole episodes to success | episodes to success: 470 -> 405 episodes | Helped |
| BO-tuned hyperparameters of DDPG (BOHAMIANN) vs original DDPG | DDPG on 2-link reaching · 2-link reaching episodes to success | episodes to success: 1512 -> 1102 episodes | Helped |
| DNGO-tuned hyperparameters of DDPG instead of BOHAMIANN | DDPG on CartPole and 2-link reaching · episodes to success | episodes to success: CartPole 470 -> 507; reaching 1512 -> 1642 | Hurt |
| BO-tuned ResNet hyperparameters and architecture (BOHAMIANN) vs hand-tuned ResNet | ResNet on CIFAR-10 · CIFAR-10 classification error after retraining on full data | classification error %: 7.40 % vs 7.51 % | No clear effect |
Closest papers by meaning
- A Study of Bayesian Neural Network Surrogates for Bayesian Optimization (ICLR 2024 · similarity 0.90)
- Understanding High-Dimensional Bayesian Optimization (ICML 2025 · similarity 0.89)
- Bayesian Optimization via Continual Variational Last Layer Training (ICLR 2025 · similarity 0.89)
- Bayesian Optimisation with Unknown Hyperparameters: Regret Bounds Logarithmically Closer to Optimal (NeurIPS 2024 · similarity 0.88)
- Scalable Bayesian Optimization via Focalized Sparse Gaussian Processes (NeurIPS 2024 · similarity 0.88)
- Scalable Bayesian Learning with posteriors (ICLR 2025 · similarity 0.88)
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.