eyangch / rl-evo

Reinforcement Learning vs Evolution

Reinforcement learning and evolutionary algorithms both aim to train an agent, or robot, to accomplish some task. For example, these algorithms can be used to create robots that can play games, navigate environments, and interact with the real world. These algorithms also utilize neural networks as the structure of its "brain". Neural networks take in data (in this case usually its environment) and produce an output (in this case its next action). Adjusting the parameters, or weights, of the neural network in a certain way can result in the neural network being able to accomplish some sort of task. For a more detailed dive into how neural networks function, click on the image below.

Even though both reinforcement learning and evolutionary algorithms use neural networks, they have completely different ways of learning and using new information and data. How do they work?

Reinforcement Learning

Evolutionary Algorithms

Comparing Algorithms