Analysis of different deep Reinforcement Learning methods
random_action.py : Agent without training
- train.py : tflearn code for solving cartpole environnment without using Reinforcement learning
- deepq.py : PyTorch code for solving cartpole environnment using Deep Q-learning
- cartpole-pg-tf : Policy Gradient where policy function is a neural network (written using TensorFlow)
- cartpole-pg.py : Policy gradient where policy function is evaluated using dot product between randomly generated numbers and state of the env. Deep Learning is not used.
- a2c-cartpole.py : Advantage Actor-critic algorithm for solving the Cartpole environment (in PyTorch)