Covariance matrix adaptation is comparable on real-valued hyperparameters, but you're stuck outside of that. More typical genetic algorithms waste a TON of comp

Covariance matrix adaptation is comparable on real-valued hyperparameters, but you're stuck outside of that. More typical genetic algorithms waste a TON of compute time, so they might end up finding a good solution, but the computational budget is out of reach of normal companies.