/CS419-Project

Analysis of different deep Reinforcement Learning methods

Primary LanguagePython

CS419-Project

Analysis of different deep Reinforcement Learning methods

Repository Structure:

random_action.py : Agent without training

Non-RL

  • train.py : tflearn code for solving cartpole environnment without using Reinforcement learning

Deep_Qlearn

  • deepq.py : PyTorch code for solving cartpole environnment using Deep Q-learning

PolicyGradient

  • 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.

Actor-Critic

  • a2c-cartpole.py : Advantage Actor-critic algorithm for solving the Cartpole environment (in PyTorch)