REMBO

Overview

REMBO implements Bayesian optimization for high-dimensional problems by using random embeddings to reduce the effective search space. It maps the original high-dimensional space into a lower-dimensional subspace where standard Bayesian optimization is performed.

Best for

Best for
Researchers and engineers working in Matlab who need to optimize high-dimensional black-box functions with limited evaluations.

Use cases

• Optimizing hyperparameters for machine learning models with many parameters
• Tuning complex simulation or engineering design parameters
• Performing black-box optimization when function evaluations are expensive

Notes

REMBO implements Bayesian optimization for high-dimensional problems by using random embeddings to reduce the effective search space. It maps the original high-dimensional space into a lower-dimensional subspace where standard Bayesian optimization is performed.

116 stars on GitHub. Last updated 2013-08-04.

Use cases

• Optimizing hyperparameters for machine learning models with many parameters
• Tuning complex simulation or engineering design parameters
• Performing black-box optimization when function evaluations are expensive

Pros

• Handles high-dimensional optimization where standard Bayesian optimization fails
• Backed by theoretical guarantees on the embedding approach
• Lightweight and focused implementation in Matlab

Cons

• Limited to Matlab, reducing accessibility for Python-heavy workflows
• Small community with 116 stars and minimal recent updates
• May require tuning of embedding dimension for best results

Indexed from awesome-llmops and enriched against its public facts.