gradient algorithm (Advantage Actor Critic - A2C) and an evolutionary algorithm (ES) for the cartpole problem on OpenAI gym.Using Q learning we train a state space model within the environment. We look at the CartPole reinforcement learning problem. Introduction to Reinforcement Learning - Cartpole DQN.I wish I can solve it in 2000 episodes so that is my outer loop. In the challenge, we want to keep the pole on the cart as long as possible. An episode is like a round in typical video action-fighting games. This one works on an environment named CartPole-v0. The scaffold of a gym challenge is to first build the environment.Deep Q Network (DQN) and its extensions (Double-DQN, Dueling-DQN, Prioritized Experience Replay). While I used state signals from Gym directly, this might also work for the images. Double DQN yields a much lower variance and a better policy from my experiments on CartPole. DQN is known to have a very high variance and there is no guarantee of convergence. The pole is unstable and tends to fall over. This environment contains a wheeled cart balancing a vertical pole. We’ll be using OpenAI Gym to provide the environments for learning.
