Skip to content
No description, website, or topics provided.
Python Jupyter Notebook
Branch: master
Clone or download
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
assets add: Basic Actor Critic (A2C) (#11) Apr 9, 2017
tests add: test codes (#7) Apr 2, 2017
utils [fix] env has to be reset before calling env.step(action) Apr 8, 2018
.gitignore feat: Gym uploader (#9) Apr 3, 2017
01_0_play_frozenlake_det.py Initial commit Mar 3, 2017
01_1_play_frozenlake_det.py Added gitignore Mar 3, 2017
02_random_frozenkake_det.py Initial commit Mar 3, 2017
03_0_q_table_frozenlake_det.py
03_1_q_table_frozenlake_det.py
03_2_q_table_frozenlake_det.py Initial commit Mar 3, 2017
04_play_frozenlake.py [fix] env has to be reset before calling env.step(action) Apr 8, 2018
05_0_q_table_frozenlake.py Initial commit Mar 3, 2017
05_q_table_frozenlake.py [fix] Update hyperparameters for better performance & fix test run Apr 9, 2018
06_q_net_frozenlake.py Initial commit Mar 3, 2017
07_0_random_cartpole.py Initial commit Mar 3, 2017
07_1_q_net_cartpole.py Initial commit Mar 3, 2017
07_2_dqn_2013_cartpole.py refactor: DQN docstring & fix divergence (#21) Jun 6, 2017
07_3_dqn_2015_cartpole.py refactor: DQN docstring & fix divergence (#21) Jun 6, 2017
08_1_pg_cartpole.py add: reward normalization Apr 1, 2017
08_2_softmax_pg_cartpole.py add: reward normalization Apr 1, 2017
08_3_softmax_pg_pacman.py
08_4_softmax_pg_pong.py log 0 -> nan problem (#16) Apr 30, 2017
08_4_softmax_pg_pong_y.py
09_2_cross_entropy.py feat: Cross entropy method (#10) Apr 4, 2017
10_1_Actor_Critic.ipynb Update 10_1_Actor_Critic.ipynb Apr 20, 2018
10_2_A3C_threads.py add: simple A3C implementations (#18) Jun 4, 2017
README.md
dqn.py refactor: DQN docstring & fix divergence (#21) Jun 6, 2017
gym.ini feat: Gym uploader (#9) Apr 3, 2017
gym_uploader.py feat: Gym uploader (#9) Apr 3, 2017
mini_pacman.py Added buggy lab-08 code Mar 3, 2017
requirements.txt Added requirements.txt Mar 3, 2017

README.md

Reinforcement Zero to All

This is work in progress and it may have bugs. However, we call for your comments and pull requests.

We emphasize on the following:

  • Readiability over anything else
    • That's why we choose Python
  • Pythonic code
    • PEP8
    • Docstring
  • Use High Level Tensorflow API
    • Cleaner and easier to understand
  • KISS

Lecture videos

File naming rule

99_9_description.py
  • First two digits indicates a category of algorithms
    • 07: DQN
    • 08: Policy Gradient
    • 09: Random Search Methods
    • 10: Actor Critic
  • A second digit indicates an id
  • Description shows what the file is about

How to use uploader

It makes the uploading process a little bit simpler

  1. Go to https://gym.openai.com/
  2. Login with your github account
  3. Copy your OpenAI api key from the upper right corner of your profile page
    user
  4. Modify gym.ini
  5. In console
#python gym_uploader.py /path/to/gym_results
python gym_uploader.py gym-results/

Install requirements

pip install -r requirements.txt

Run test and autopep8

TODO: Need to add more test cases

pytest
# pip install autopep8 # if you haven't install
autopep8 . --recursive --in-place --pep8-passes 2000 --verbose --ignore E501

Contributions/Comments

We always welcome your comments and pull requests.

You can’t perform that action at this time.