Evolutionary Algorithms
An evolutionary algorithm (EA) is an algorithm that uses mechanisms inspired by nature and solves problems through processes that emulate the behaviors of living organisms.
Topics
How is it different from RL?
Evolutionary strategies (ES) or genetic algorithms search directly in the space of policy parameters. They treat the policy as a black box, optimizing it based only on how well it performs (its cumulative reward).
RL has a structure that can be exploited. We optimize with respect to gradients.
ES simply treats things as black boxes, only keeping the best ones.
RL methods leverage every piece of data (state, action, reward, next state): They update the policy or value function after each episode or even after each step.