NikhilDange1 / Cartpole-DQN-agent

A Reinforcement learning based Deep Q network agent implemented in python using Keras

Cartpole-DQN-agent

A Deep Reinforcement Learningn agent implemented in python using Keras with TensorFlow as backend. This algorithm can be found in the Playing Atari with Deep Reinforcement Learning paper

Prerequisite

The code runs on Python 3.6 and uses the following modules

• Numpy
• Keras
• Tensorflow(backend)
• OpenAI gym

Modules can be installed using the requirements.txt file

Files

• DQNagent.py-The file contains the Deep Q netwrok agent object.

• Env.py-This file runs the OpenAI gym simulation and calls the DQN class.

Hyperparameters

agent.memory(100000)
batch = 64 #trainig batch size
eph = 0.9 #ephsilon
eph_min = 0.01
decay = 0.995 #decay for ephsilon
gamma = 0.99 # discount factor

Result

Cartpole environment