I use GA optimization for non-convex problems, mainly hyperparameter optimization. Sometimes it’s very effective but I’ve not found a way to know ahead of time if it will outperform other algorithms.