[KP] Towards a Benchmark for EBBO. Enabling Feature-model-based Algorithm Comparison

Benchmarking and algorithm comparison is an essential part of algorithm development process, especially in cases when performance or quality of the result is important. Expensive blackbox optimization (EBBO) algorithms fall into this category.
BRISE-MPL is a multiple product line for EBBO, developed at the chair of software technology. Its main feature is a highly-customizable optimization workflow, enabling high solution quality and ease of extensibility. BRISE can also serve as a benchmark, enabling comparison of newly developed approaches.
In its current state, BRISE-based benchmark analysis is manually created for each use case, resulting into weak maintainability and reuse. The goal of this internship is to enable feature-model-based automatic creation of benchmark analysis.
To reach this goal the following tasks need to be accomplished

• Analyze state-of-the-art benchmarks for EBBO, including BRISE benchmarking module.
• Derive a feature model and a set of software artifacts.
• Implement the proposed approach, including extension of the test suite.
• Integrate the module into the latest release version of BRISE.

Betreuer: Dmytro Pukhkaiev