Genetic algorithms essentially run coordinate descent with some tweaks. On the positive side, coordinate descent doesn't need gradients and thus happily optimiz

Genetic algorithms essentially run coordinate descent with some tweaks. On the positive side, coordinate descent doesn't need gradients and thus happily optimizes complex models. On the negative side, coordinate descent is roughly N times slower than gradient descent where N is the number of parameters being optimized.