You can use genetic algorithms for training the weights of a neurL network. This is called neuroevolution and it's competitive on reinforcement learning because

You can use genetic algorithms for training the weights of a neurL network. This is called neuroevolution and it's competitive on reinforcement learning because gradients are harder to find there...