Research map / Probabilistic and scientific modeling / Bayesian optimization
Taking the GP Out of the Loop
Mehul Bafna, Siddhant Anand Jadhav, David Sweet
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 |
|---|---|---|---|
| replace GP surrogate with ENN and Thompson sampling with UCB | TuRBO · Hopper-v5 (frozen noise, N=50000) | total proposal time: 5532.4 -> 41.2 | Improved |
| replace GP surrogate with ENN | TuRBO · LunarLander-v3 (natural noise) | total proposal time: 2128.5 -> 249.3 | Improved |
| replace GP surrogate with ENN | TuRBO · LASSO-DNA (180D) | total proposal time: 6508.1 -> 108.1 | Improved |
| replace GP surrogate with ENN | TuRBO · several simulators | solution quality | No clear effect |
| omit hyperparameter fitting and use non-dominated sort for noise-free problems | TuRBO-ENN · frozen-noise problems | proposal time: speedups up to about 150x vs 8.5x to 80x natural | Improved |
| remove surrogate, sample TuRBO candidates uniformly (turbo-zero) (ablation) | TuRBO-ENN · frozen-noise problems | optimization quality | Mixed |
| remove surrogate, sample TuRBO candidates uniformly (turbo-zero) (ablation) | TuRBO-ENN · natural-noise problems | optimization quality | Hurt |
| use ENN surrogate instead of exact GP, fit time | surrogate fitting · Ackley D=10 pure fitting | fit time | Improved |
| use ENN surrogate instead of other surrogates, test log-likelihood | surrogate fitting · Ackley D=10 pure fitting | test log-likelihood rank | No clear effect |
| use ENN instead of Vecchia GP nearest-neighbor method | TuRBO · proposal time vs N | proposal time | Improved |
Closest papers by meaning
- A Study of Bayesian Neural Network Surrogates for Bayesian Optimization (ICLR 2024 · similarity 0.89)
- Bayesian Optimisation with Unknown Hyperparameters: Regret Bounds Logarithmically Closer to Optimal (NeurIPS 2024 · similarity 0.88)
- Turbocharging Gaussian Process Inference with Approximate Sketch-and-Project (NeurIPS 2025 · similarity 0.88)
- Bayesian Optimization via Continual Variational Last Layer Training (ICLR 2025 · similarity 0.87)
- Understanding High-Dimensional Bayesian Optimization (ICML 2025 · similarity 0.87)
- Standard Gaussian Process is All You Need for High-Dimensional Bayesian Optimization (ICLR 2025 · similarity 0.87)
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.
