hrnoh / cartpole-dqn Goto Github PK

Simulation for cart-pole balancing problem

* Usage : Run main.m

* Stable Training Episode number : about 8000~

* Check Neural Network Performance : 
  - Changing obj.net.trainParam.showWindow = false -> true (32 lines) in QNetwork.m file.

-------------------------------------------------
1. main.m
    - A script for Cart-Pole balancing demo
2. HParams.m
    - Hyper parameters class
3. DQNLearner.m
    - A class that learn Deep Q Networks
4. Environment.m
    - Cart-pole environment class (created trainig data sets automatically)
5. QNetwork.m
    - Q Network class
6. ReplayMemory.m
    - A class that stores transitions
7. drawcartpend.m
    - plot cart-pole function