README.md

temporally_correlated_exploration_stochasticity

Code for the poster presentation at Japanese Neural Network Society Meeting 2018.

Please check the (non-peer-reviewed) conference paper (Improving exploration in reinforcement learning with temporally correlated stochasticity.pdf) and poster (poster.pdf) for details.

Required Libraries:

• TensorFlow
• OpenAI Gym
• scipy.io

Explanation

Please prepare a data folder '..\data'. By running python programs, the result will be saved into the data folder, as two .mat files, for Gaussian white exploration noise and OU exploration noise .

Run the code

python cartpole.py

or

python chain_world.py

for getting the results in Fig.2 in the paper.