Competition overview

This competition invites researchers to test their global optimization algorithms against a meticulously curated set of 24 problem instances from the Generalized Numerical Benchmark Generator (GNBG). The GNBG-III competition introduces the next generation of property-aware and computationally hard numerical benchmarks, designed to expose the behavior of global optimization algorithms under controlled structural conditions. The suite encompasses:

This competition presents problems with diverse characteristics such as modality, ruggedness, asymmetry, conditioning, and deceptiveness, providing a thorough test of algorithmic performance. Beyond solution quality, the emphasis is on understanding how algorithms reach solutions. Participants will explore how algorithms handle deceptive landscapes, traverse valleys, and adapt to varying problem difficulties, offering deeper insight into optimization in complex numerical environments.