IOHexperimenter: Benchmarking Platform for Iterative Optimization Heuristics

Area of Science:

• Computational intelligence
• Algorithm analysis
• Optimization techniques

Background:

• Benchmarking iterative optimization heuristics is crucial for performance evaluation.
• Existing tools may lack flexibility or detailed logging capabilities.
• The IOHprofiler project aims to provide a comprehensive environment for empirical algorithm analysis.

Purpose of the Study:

• Introduce IOHexperimenter, the experimentation module of the IOHprofiler project.
• Provide a user-friendly and customizable toolbox for benchmarking various iterative optimization heuristics.
• Facilitate efficient performance analysis and comparison of optimization algorithms.

Main Methods:

• IOHexperimenter offers an environment for creating customized problem suites.
• It provides granular logging options for the optimization process.
• The module ensures compatibility with existing data analysis tools for streamlined benchmarking pipelines.

Main Results:

• IOHexperimenter enables easy benchmarking of algorithms like local search, evolutionary algorithms, and Bayesian optimization.
• It provides an efficient interface between optimization problems and their solvers.
• Granular logging significantly speeds up the deployment of benchmarking pipelines.

Conclusions:

• IOHexperimenter is a valuable and flexible tool for researchers and practitioners in iterative optimization.
• Its customizable nature and detailed logging enhance the efficiency and depth of algorithm benchmarking.
• The module integrates seamlessly with the IOHprofiler environment, promoting reproducible research.