mlrMBO: R Toolbox for Bayesian Optimization of Black-Box Functions

mlrMBO is a flexible R toolbox for model-based optimization, also known as Bayesian optimization, authored by Bernd Bischl. It implements the Efficient Global Optimization Algorithm for single- and multi-objective problems with mixed continuous, categorical, and conditional parameters. The toolbox integrates with the 'mlr' machine learning library for regression modeling and provides features for parallel execution, visualization, and logging.

Use Cases

• Hyperparameter tuning for machine learning models based on the support for mixed parameter types and infill criteria.
• Multi-objective optimization of engineering or simulation models based on the described multi-objective capabilities.
• Teaching and understanding algorithm behavior based on the visualization and sophisticated logging mechanisms mentioned.
• Parallel batch optimization of expensive functions based on the multi-point batch proposal and parallel execution features.

Strengths

• Implements the Efficient Global Optimization Algorithm, a recognized method in the field.
• Designed for both single- and multi-objective optimization with mixed parameter types.
• Integrates with the 'mlr' toolbox, which provides dozens of regression learners for modeling.
• Features a modular implementation allowing users to replace or adapt components.

Limitations

• Row count and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

Bernd Bischl

Collection Method

Software toolbox implementation.