Eventually, all you need is a simple evolutionary algorithm (for neuroevolution of continuous control policies).

Michel El Saliby, Giorgia Nadizar, Erica Salvato, Eric Medvet

Browse the full GECCO paper archive.