Abstract
Multi-objective optimisation is a prominent subfield of optimisation with high relevance in real-world problems, such as engineering design. Over the past 2 decades, a multitude of heuristic algorithms for multi-objective optimisation have been introduced and some of them have become extremely popular. Some of the most promising and versatile algorithms have been implemented in software platforms. This article experimentally investigates the process of interpreting and implementing algorithms by examining multiple popular implementations of three well-known algorithms for multi-objective optimisation. We observed that official and broadly employed software platforms interpreted and thus implemented the same heuristic search algorithm differently. These different interpretations affect the algorithmic structure as well as the software implementation. Numerical results show that these differences cause statistically significant differences in performance.