Benchmarking Beyond Statistics - Data-Driven Footprints for Explainable Black-Box Optimization

Explore emerging benchmarking and meta-learning methodologies that are transforming the evaluation and selection of optimization algorithms in this 31-minute seminar talk. Discover how the field is moving beyond traditional statistical approaches toward trustworthy and explainable paradigms through two key innovations. Learn about representative instance selection techniques that ensure benchmarking data remains diverse and generalizable rather than being limited to narrow test sets. Understand the revolutionary concept of algorithmic footprints—digital signatures that capture how algorithms interact with problem landscapes and reveal which landscape features determine their success or failure. Examine how these developments are creating a new generation of explainable and automated optimization systems. Gain insights into how interpretable, data-grounded approaches are replacing simple statistics to advance black-box optimization toward greater transparency, reproducibility, and knowledge transferability.